Original Research

Spectators at baseball games may be exposed to excess solar UV radiation (UVR), which has been linked to the development of both melanoma and nonmelanoma skin cancers.^1,2 Although baseball hats traditionally are worn to demonstrate team support, they also may provide some sun protection for the head and face where skin cancers are commonly found.

The importance of protecting the skin from solar UVR has led to sun-protection programs and community education as well as efforts to evaluate the impact of these programs. Major League Baseball (MLB) has partnered with the American Academy of Dermatology since 1999 to promote the importance of sun protection and raise skin cancer awareness through its Play Sun Smart program.³ A study conducted 10 years ago (N=2030) evaluated hat use in spectators at MLB games and noted that less than half of all spectators in seating sections exposed to direct sunlight wore hats.⁴ The purpose of the current study was to evaluate how public education about sun protection has impacted the use of hats by spectators at MLB games in 2015 compared to the prior study in 2006.

Methods

Data were collected during a 3-game series (2 day games, 1 night game) in August 2015 in New York, New York. During one of the day games, 18,000 fans received a free wide-brimmed hat. High-resolution digital photographs of seating sections were obtained using a camera with a 300-mm lens. Using the same methodology as the prior study,⁴ sunny and shaded seating sections were photographed during all 3 games (Figure). Photographs of each section were analyzed by an independent reviewer using a high-resolution computer screen. Spectators wearing head coverings—baseball hats, visors, or hats with circumferential brims—were defined as using hats. The number of spectators wearing hats versus not wearing hats was recorded for all identical sections of interest. Bleacher seating was analyzed separately, as spectators presumably knew in advance of the continuous direct sun exposure during day games, and a subset of young children in the bleachers (<10 years of age) also was assessed. A continuously sunny section also was evaluated at the second and sixth innings to see if hats were presumably purchased during exposure. Statistical significance was determined using χ² tests with P<.05 indicating statistical significance.

Results

This analysis consisted of 3539 spectators. In both the sunny and shaded sections of a day game, there were more spectators wearing hats (49% and 37%, respectively)(P<.001) than in the same sections at night games (35% and 29%, respectively)(Table 1). During the day game, more spectators wore hats in the sunny section than in the adjacent shaded section (49% vs 37%; P<.001). Analysis of the same 2 sections during the night game revealed no significant differences.

Spectators sitting in the bleachers during a day game who presumably knew to anticipate direct sun exposure showed no significant differences in hat-wearing patterns versus the sunny section (44% vs 49%) but were more likely to wear hats compared to those sitting in the bleachers at the night game (44% vs 33%)(P<.001)(Table 1). There was no significant difference in the number of hats worn by spectators in the sunny section in the second inning (43%) versus the same section after continuous sun exposure at the sixth inning (44%)(Table 2). Significantly more children seated in the bleachers during the day game wore hats compared to adults in the same section (64% vs 42%; P<.001)(Table 3). During the hat giveaway day, significantly more spectators wore hats (the majority of which were the free giveaway hats) across all sections studied (P<.001)(Table 4).

Comment

More than 23 million spectators attended daytime MLB games in 2015, with millions more attending minor league and amateur events.⁵Although sun-protection messages tend to be well understood and received by society, many choose to ignore them.⁶

In partnership with the American Academy of Dermatology, the MLB’s Play Sun Smart program has promoted UVR risk awareness at sporting events since 1999.³ Those affiliated with MLB teams also receive annual skin cancer screenings in conjunction with a public education effort in May of each season. However, despite the years of sun-protection education, our study found that less than half of attendees wore hats for UVR protection. In fact, there were no significant differences noted across all of the hat-wearing parameters studied (day vs night game, sunny vs shaded section, sunny section over course of game) between the current study compared to the results from 10 years prior⁴ (Tables 1 and 2). For spectators in the bleacher section, even presumably knowing in advance that seating would be in the sun did not significantly increase hat-wearing behavior. Although skin cancer rates continue to rise, hat-wearing trends remain stable, revealing a concerning trend.

Increased availability of sunscreen has led to improved sun-protective behaviors in many populations.⁷ In our study, the free hat giveaway had the greatest impact on hat wearing, which suggests that improved availability and access to hats can lead to an important opportunity for sun-protection programs to partner with hat manufacturers to augment their use and protective impact.

Sun avoidance during childhood and adolescence has been shown to decrease the risk for melanoma.¹ Young children had the highest rate of hat usage in the current study, possibly due to parental example or dictates. Research has shown the importance of role models in promoting sun safety to young children,^8,9 so perhaps use of hats by parents or MLB players contributed to the hat-wearing behavior observed in this subpopulation.

Given the limited change observed in hat-wearing behaviors over the last decade, a knowledge and behavioral gap appears to exist that may be able to be exploited to enhance future sun protection. Also, based on our findings, the MLB and other sun-protection education campaigns may wish to augment their UVR protective messages by offering hat giveaways, which appear to have a notable impact.

Acknowledgment

The authors thank Jessie Skapik, BS (New York, New York), for her independent review of the spectator photographs.

Spectators at baseball games may be exposed to excess solar UV radiation (UVR), which has been linked to the development of both melanoma and nonmelanoma skin cancers.^1,2 Although baseball hats traditionally are worn to demonstrate team support, they also may provide some sun protection for the head and face where skin cancers are commonly found.

The importance of protecting the skin from solar UVR has led to sun-protection programs and community education as well as efforts to evaluate the impact of these programs. Major League Baseball (MLB) has partnered with the American Academy of Dermatology since 1999 to promote the importance of sun protection and raise skin cancer awareness through its Play Sun Smart program.³ A study conducted 10 years ago (N=2030) evaluated hat use in spectators at MLB games and noted that less than half of all spectators in seating sections exposed to direct sunlight wore hats.⁴ The purpose of the current study was to evaluate how public education about sun protection has impacted the use of hats by spectators at MLB games in 2015 compared to the prior study in 2006.

Methods

Data were collected during a 3-game series (2 day games, 1 night game) in August 2015 in New York, New York. During one of the day games, 18,000 fans received a free wide-brimmed hat. High-resolution digital photographs of seating sections were obtained using a camera with a 300-mm lens. Using the same methodology as the prior study,⁴ sunny and shaded seating sections were photographed during all 3 games (Figure). Photographs of each section were analyzed by an independent reviewer using a high-resolution computer screen. Spectators wearing head coverings—baseball hats, visors, or hats with circumferential brims—were defined as using hats. The number of spectators wearing hats versus not wearing hats was recorded for all identical sections of interest. Bleacher seating was analyzed separately, as spectators presumably knew in advance of the continuous direct sun exposure during day games, and a subset of young children in the bleachers (<10 years of age) also was assessed. A continuously sunny section also was evaluated at the second and sixth innings to see if hats were presumably purchased during exposure. Statistical significance was determined using χ² tests with P<.05 indicating statistical significance.

Results

This analysis consisted of 3539 spectators. In both the sunny and shaded sections of a day game, there were more spectators wearing hats (49% and 37%, respectively)(P<.001) than in the same sections at night games (35% and 29%, respectively)(Table 1). During the day game, more spectators wore hats in the sunny section than in the adjacent shaded section (49% vs 37%; P<.001). Analysis of the same 2 sections during the night game revealed no significant differences.

Spectators sitting in the bleachers during a day game who presumably knew to anticipate direct sun exposure showed no significant differences in hat-wearing patterns versus the sunny section (44% vs 49%) but were more likely to wear hats compared to those sitting in the bleachers at the night game (44% vs 33%)(P<.001)(Table 1). There was no significant difference in the number of hats worn by spectators in the sunny section in the second inning (43%) versus the same section after continuous sun exposure at the sixth inning (44%)(Table 2). Significantly more children seated in the bleachers during the day game wore hats compared to adults in the same section (64% vs 42%; P<.001)(Table 3). During the hat giveaway day, significantly more spectators wore hats (the majority of which were the free giveaway hats) across all sections studied (P<.001)(Table 4).

Comment

More than 23 million spectators attended daytime MLB games in 2015, with millions more attending minor league and amateur events.⁵Although sun-protection messages tend to be well understood and received by society, many choose to ignore them.⁶

In partnership with the American Academy of Dermatology, the MLB’s Play Sun Smart program has promoted UVR risk awareness at sporting events since 1999.³ Those affiliated with MLB teams also receive annual skin cancer screenings in conjunction with a public education effort in May of each season. However, despite the years of sun-protection education, our study found that less than half of attendees wore hats for UVR protection. In fact, there were no significant differences noted across all of the hat-wearing parameters studied (day vs night game, sunny vs shaded section, sunny section over course of game) between the current study compared to the results from 10 years prior⁴ (Tables 1 and 2). For spectators in the bleacher section, even presumably knowing in advance that seating would be in the sun did not significantly increase hat-wearing behavior. Although skin cancer rates continue to rise, hat-wearing trends remain stable, revealing a concerning trend.

Increased availability of sunscreen has led to improved sun-protective behaviors in many populations.⁷ In our study, the free hat giveaway had the greatest impact on hat wearing, which suggests that improved availability and access to hats can lead to an important opportunity for sun-protection programs to partner with hat manufacturers to augment their use and protective impact.

Sun avoidance during childhood and adolescence has been shown to decrease the risk for melanoma.¹ Young children had the highest rate of hat usage in the current study, possibly due to parental example or dictates. Research has shown the importance of role models in promoting sun safety to young children,^8,9 so perhaps use of hats by parents or MLB players contributed to the hat-wearing behavior observed in this subpopulation.

Given the limited change observed in hat-wearing behaviors over the last decade, a knowledge and behavioral gap appears to exist that may be able to be exploited to enhance future sun protection. Also, based on our findings, the MLB and other sun-protection education campaigns may wish to augment their UVR protective messages by offering hat giveaways, which appear to have a notable impact.

Acknowledgment

The authors thank Jessie Skapik, BS (New York, New York), for her independent review of the spectator photographs.

Spectators at baseball games may be exposed to excess solar UV radiation (UVR), which has been linked to the development of both melanoma and nonmelanoma skin cancers.^1,2 Although baseball hats traditionally are worn to demonstrate team support, they also may provide some sun protection for the head and face where skin cancers are commonly found.

The importance of protecting the skin from solar UVR has led to sun-protection programs and community education as well as efforts to evaluate the impact of these programs. Major League Baseball (MLB) has partnered with the American Academy of Dermatology since 1999 to promote the importance of sun protection and raise skin cancer awareness through its Play Sun Smart program.³ A study conducted 10 years ago (N=2030) evaluated hat use in spectators at MLB games and noted that less than half of all spectators in seating sections exposed to direct sunlight wore hats.⁴ The purpose of the current study was to evaluate how public education about sun protection has impacted the use of hats by spectators at MLB games in 2015 compared to the prior study in 2006.

Methods

Data were collected during a 3-game series (2 day games, 1 night game) in August 2015 in New York, New York. During one of the day games, 18,000 fans received a free wide-brimmed hat. High-resolution digital photographs of seating sections were obtained using a camera with a 300-mm lens. Using the same methodology as the prior study,⁴ sunny and shaded seating sections were photographed during all 3 games (Figure). Photographs of each section were analyzed by an independent reviewer using a high-resolution computer screen. Spectators wearing head coverings—baseball hats, visors, or hats with circumferential brims—were defined as using hats. The number of spectators wearing hats versus not wearing hats was recorded for all identical sections of interest. Bleacher seating was analyzed separately, as spectators presumably knew in advance of the continuous direct sun exposure during day games, and a subset of young children in the bleachers (<10 years of age) also was assessed. A continuously sunny section also was evaluated at the second and sixth innings to see if hats were presumably purchased during exposure. Statistical significance was determined using χ² tests with P<.05 indicating statistical significance.

Results

This analysis consisted of 3539 spectators. In both the sunny and shaded sections of a day game, there were more spectators wearing hats (49% and 37%, respectively)(P<.001) than in the same sections at night games (35% and 29%, respectively)(Table 1). During the day game, more spectators wore hats in the sunny section than in the adjacent shaded section (49% vs 37%; P<.001). Analysis of the same 2 sections during the night game revealed no significant differences.

Spectators sitting in the bleachers during a day game who presumably knew to anticipate direct sun exposure showed no significant differences in hat-wearing patterns versus the sunny section (44% vs 49%) but were more likely to wear hats compared to those sitting in the bleachers at the night game (44% vs 33%)(P<.001)(Table 1). There was no significant difference in the number of hats worn by spectators in the sunny section in the second inning (43%) versus the same section after continuous sun exposure at the sixth inning (44%)(Table 2). Significantly more children seated in the bleachers during the day game wore hats compared to adults in the same section (64% vs 42%; P<.001)(Table 3). During the hat giveaway day, significantly more spectators wore hats (the majority of which were the free giveaway hats) across all sections studied (P<.001)(Table 4).

Comment

More than 23 million spectators attended daytime MLB games in 2015, with millions more attending minor league and amateur events.⁵Although sun-protection messages tend to be well understood and received by society, many choose to ignore them.⁶

In partnership with the American Academy of Dermatology, the MLB’s Play Sun Smart program has promoted UVR risk awareness at sporting events since 1999.³ Those affiliated with MLB teams also receive annual skin cancer screenings in conjunction with a public education effort in May of each season. However, despite the years of sun-protection education, our study found that less than half of attendees wore hats for UVR protection. In fact, there were no significant differences noted across all of the hat-wearing parameters studied (day vs night game, sunny vs shaded section, sunny section over course of game) between the current study compared to the results from 10 years prior⁴ (Tables 1 and 2). For spectators in the bleacher section, even presumably knowing in advance that seating would be in the sun did not significantly increase hat-wearing behavior. Although skin cancer rates continue to rise, hat-wearing trends remain stable, revealing a concerning trend.

Increased availability of sunscreen has led to improved sun-protective behaviors in many populations.⁷ In our study, the free hat giveaway had the greatest impact on hat wearing, which suggests that improved availability and access to hats can lead to an important opportunity for sun-protection programs to partner with hat manufacturers to augment their use and protective impact.

Sun avoidance during childhood and adolescence has been shown to decrease the risk for melanoma.¹ Young children had the highest rate of hat usage in the current study, possibly due to parental example or dictates. Research has shown the importance of role models in promoting sun safety to young children,^8,9 so perhaps use of hats by parents or MLB players contributed to the hat-wearing behavior observed in this subpopulation.

Given the limited change observed in hat-wearing behaviors over the last decade, a knowledge and behavioral gap appears to exist that may be able to be exploited to enhance future sun protection. Also, based on our findings, the MLB and other sun-protection education campaigns may wish to augment their UVR protective messages by offering hat giveaways, which appear to have a notable impact.

Acknowledgment

The authors thank Jessie Skapik, BS (New York, New York), for her independent review of the spectator photographs.

In the U.S., 4.5 million ambulatory care visits occur annually due to adverse drug reactions (ADRs) of prescription medications.¹ Many ADRs are severe, and they result in more than 100,000 death per year.² A significant number of these ADRs are preventable and are a result of inappropriate prescribing.³ It is well-documented that inappropriate prescribing is exacerbated by the number of patients who see multiple prescribers and use many different prescription medications.⁴ This situation results in many versions of a patient’s medication list and in discrepancies across data sources.⁵

Medication list discrepancies have been researched in the context of care transitions between the hospital and home.^6,7 However, less attention has been given to community-dwelling adults who use multiple outpatient prescribers, a practice common among older adults with chronic conditions who see a primary care provider and several specialists.⁴ Also, veterans are a growing patient population who use providers from multiple health care systems.⁸ Up to 70% of veterans enrolled in VA health care use both VA and non-VA providers. These patients are referred to as dual users.^9,10

There has been an increasing push for patients to be more actively engaged in their own health care, including maintenance of their medication list and other personal health information.^11-13 Providers have realized that patients have important experiences and preferences to share about how they use medications at home.^14,15 Research suggests that patient interest and ability to use patient portals is variable and dependent on age, technical abilities, health literacy, and endorsement by their providers.¹⁶ Greater patient engagement in the medication management process is potentially advantageous, especially because providers from different health care systems often lack the capability to share medication list information.^12,17

My HealtheVet, the VA’s patient portal, offers veterans several features. For example, users can securely message providers, refill prescriptions, check appointments, self-enter information, and download their VA health record (including medication history) using the Blue Button (BB) feature. The BB is managed by the HHS to provide consistency across electronic health record platforms.^18,19

This BB medication list gives VA patients the tool they need to inform their providers about the medications they take, particularly dual users. VA patients that use multiple prescribers are subject to medication list discrepancies because of the fragmentation of information.^4,20

Objectives

The objectives of this study were to (1) describe discrepancies between VA medication lists and non-VA provider medication lists for a group of veteran dual users; (2) identify therapeutic duplications in these lists; and (3) contextualize discrepancies by interviewing non-VA providers about their medication reconciliation processes and management of dual use patients.

Methods

This analysis is based on data collected as part of a pilot randomized controlled trial by Turvey and colleagues.²¹ Veterans with a diagnosis of ≥ 1 chronic health condition (eg, diabetes, hypertension) were invited by letter to participate in a study about using online management of their health information. Interested patients were screened to meet additional inclusion criteria, such as taking ≥ 5 medications, receiving care from a non-VA provider, an appointment with a non-VA provider within the study time frame, and access to a computer, online access, and printer. 

Eligible veterans were randomized to receive either (1) BB training (intervention group) instructing patients to download the Continuity of Care Document and bring it to their non-VA provider visit; or (2) a training evaluating medical information online (control group). Training information was mailed, including written materials and phone support, to both groups. The intervention group could also access an online training link.

One of the objectives was to test whether downloading and bringing the health information to a non-VA appointment decreased medication list discrepancies. The sample was small, and differences in discrepancy rates between groups were not significant. Therefore, groups were combined for the present analysis. Visits occurred between December 2013 and December 2014. Greater detail about study design and primary results are available in the study by Turvey and colleagues.²¹

Study procedures were approved by the University of Iowa Institutional Review Board and the Iowa City VA Health Care System Research and Development Committee. All participants provided consent.

Identifying Discrepancies

A 4-phase process was used to address medication discrepancies.^22,23 The first phase defined medication discrepancy categories. The mutually exclusive categories were dose, frequency, and missing discrepancies. In cases where a medication was both dose and frequency discrepant, only dose discrepancy was applied. For missing medications, entities on only the VA list were marked as “non-VA missing” and medications appearing on only the non-VA list would be denoted as “VA missing.” Medications with no discrepancy were marked as such.

Phase 2 involved collecting medication data. Medication lists from the VA medical record were printed at the time of the non-VA provider appointment. Non-VA medication lists were obtained by sending a medical record request for the visit note, medication list, and any associated visit test results to the non-VA provider office within 2 to 3 weeks of the appointment. Patient names from both lists were replaced with unique patient identifiers.

In phase 3, a research assistant abstracted the hard copy medication lists into a database and identified discrepancies. Variables included medication name, dose, frequency, and administration route. Although administration routes were collected, discrepancies were not assessed because this information commonly was not specified. Medications also were coded as prescription or over-the-counter (OTC). Durable medical equipment often was present on VA lists (eg, syringes, test strips) and was excluded from all analyses. Medications also were not coded as discrepant if they were referenced in a visit note as being changed by the non-VA provider. These combined lists were evaluated by the research assistant based on the discrepancy categories specified in phase 1 and were verified by a pharmacist.

Phase 4 involved counting medication discrepancies. Medication discrepancy rates were calculated at the patient level, both descriptively (mean number of discrepancies per patient) and as a proportion of medications discrepant (number of discrepancies divided by total medications).

Identifying Duplications and High-Risk Medications

A pharmacist examined each combined medication list to identify therapeutic duplications, defined as a patient using ≥ 2 medications from the same medication class (eg, patient taking 2 statin drugs) but not 2 drugs for the same condition (eg, fish oil and atorvastatin for dyslipidemia). High-risk medications also were noted, including anticoagulants, certain nonsteroidal anti-inflammatory drugs, oral and injectable hypoglycemics, opioids, sedatives, and hypnotics.^24-26 These medications received special focus because of their link to a high risk for ADRs.²⁷

Descriptive statistics were calculated for patient characteristics and for each discrepancy type, both overall and according to prescription OTC, and high-risk medications. The proportion of discrepant medications was calculated for each category. Bivariate correlations were calculated for select variables to understand potential relationships.

Interviews With Non-VA Providers

All patients were instructed to bring a consent letter and the 1-page questionnaire to their non-VA provider appointment. The questionnaire contained an item asking whether non-VA providers could be contacted for a 15- to 30-minute follow-up interview. The semistructured, qualitative interviews assessed their experiences working with VA providers and VA patients, experiences with VA documents or records, preferences for receiving information from the VA, experience with personal health records, and sharing information with the VA. Eight interviews were conducted, audio-recorded, and transcribed. The goal of the interviews was to explore and understand provider perspectives on managing dual use veterans, including medication reconciliation processes to add context to the interpretation of medication list analysis. Because the data set was relatively small, summaries of each interview were created to highlight main points. These main points were sorted into topics, summarized, and representative quotes were selected.

Results

Fifty veterans were included in the analysis (Table 1). The mean age was 68.5 (SD 6.2); 90% were men. On average, they reported having 6 chronic health conditions and a fair-to-good health status. Based on the combined medication lists from VA and non-VA providers, veterans took an average of 15.8 (SD 7.0) unique medications (combined prescription and OTC/vitamins) and had an average of 10.0 (SD 6.1) all-type discrepancies (Table 2).

Overall, 58% of the prescription medications were discrepant: The most common discrepancy between the 2 lists was medication missing on one of the lists, which occurred 3.9 times per patient on average for prescription medications and 2.8 times per patient for OTCs. Frequency or dose discrepancies also were common between the lists at a rate of 1.9 discrepancies per patient for prescription medications and 1.2 discrepancies per patient for OTCs.

For high-risk medications, opiates and sedative medications had the most discrepancies between the lists because the VA practitioner may not have known that the patient was taking an opiate, although other discrepancies were present (Table 3). Anticoagulant discrepancies were the most consistent, most of these occurring with aspirin. Last, insulin commonly was dose discrepant between the 2 lists, although it also was missing from one list for a number of patients. Overall, high-risk medications shared a discrepancy rate (46.9%) similar to the overall rate.

Twelve therapeutic duplications were identified in the sample.Ten were between-list duplications, that is, “provider A” thought the patient was on a particular medication and “provider B” thought that the patient was on a different medication (Table 4). In 6 instances, within-list duplications were identified (ie, a provider had 2 medications on the list that should not be taken together because they were in the same drug class). In 4 cases, both between- and within-list duplications were present.

Interview Summaries

Veterans and medication. Multiple non-VA providers said that the primary reason veteran patients were going to a VA provider was to obtain discounted medications. The use of the VA for medications also was a way for the non-VA provider to discover that the patient was a veteran. One non-VA provider was particularly concerned about the impact of new or different medications from VA prescribers on efforts to stabilize the patient’s chronic condition.

Several non-VA providers reported that veterans often brought a medication list to the appointment, and several providers recommended the practice to their patients. Non-VA providers preferred to have patients transfer information from VA, sometimes requesting that veterans bring in their records from recent appointments rather than the non-VA provider obtain the information directly from the VA.

Information sharing. Non-VA providers generally preferred hard copies of medication lists and other documents rather than scans because they were more likely to be included in decision making if the documents were presented during the visit. Also, document scans may be buried in the electronic medical record. Some providers mentioned their interest in electronic transfer of medical information like medication lists if the technology were more developed and better integrated.

“I think the long-term vision would be that it should be electronic… it wouldn’t necessarily be feasible at this time. Our system scans paper documents in to an e-version. … but when the patient comes to their encounter 10 days later, you don’t realize the stuff’s there… Having the patient bring them in is probably a more certain way to make sure that it’s actually included in your decision making as a provider.”

Most non-VA providers welcomed more information such as imaging studies because they reported rarely receiving this information from the VA system. Two mentioned the potential for too much information and wanted concise reports should the flow of information increase. Providers had little interest in logging in to a patient’s online health record portal as a delegate for reasons related to complexity, time, privacy, and lack of mechanism to document the information accessed.

Medication reconciliation. Non-VA providers generally reported that patients bringing their own or an outside medication list would prompt a process of medication reconciliation. The providers were interested in making changes to their records based on other lists, but outside data were verified against a patient self-report of actual use before adopting changes.

“I print out my med list of what I have in the computer and then I just check off my list against their list. And then whatever’s remaining, we talk about what the differences are, when they were changed, what they were changed for, if they were taken off of something, and if I don’t agree, then I’ll tell the patient, ‘look, there’s a disagreement here, they’ve told you not to be on this. I want you on this.”

Should a discrepancy arise, non-VA providers generally had a negative view of attempting to contact VA providers. Other mechanisms such as calling a local pharmacy would be done first.

Discussion

This study provided initial evidence that medication list discrepancies exist for dual use veterans. Other studies of medication list discrepancies have linked such inconsistencies to medication-related problems and negative outcomes for patients.²⁷ Although efforts to increase access to care for veterans have advantages related to expediency, consequences to fragmenting care exist. More robust mechanisms for establishing and maintaining medication list consistency are needed, especially given the lack of a universally accepted medical record format or repository. A multifaceted approach, including patient engagement, seems necessary.

This study also showed that discrepancies of high-risk medications are common for veteran participants, placing them at risk for medication-related problems and harm. These risks included dose and frequency discrepancies that could result in over- or underdosing of medications and in medication omissions, which could cause duplicative therapies and unnecessary risks. For example, aspirin frequently was listed on non-VA lists but was omitted from VA medication lists. This could be problematic for patients who present to the VA for a procedure in which no information about aspirin could jeopardize their safety. Insulin doses also were commonly discrepant, which could impact glycemic control.

Many providers also had incomplete prescribing information for opiates. Those prescriptions are particularly relevant given the link between veterans, posttraumatic stress disorder, depression, and substance abuse.^28-30 However, it was beyond the scope of this pilot study to link these discrepancies to ADRs, such as emergency department visits or hospital admissions. Other studies have demonstrated that discrepancies at hospital discharge can result in these types of negative outcomes.^27,31 Subsequent research should determine the clinical significance of discrepancies that occur when veterans are dual users.

The qualitative interviews provided some initial context on prescriber perspectives about the role of veterans participating in the medication list sharing process and personal health records. It seemed that for the portion of patients who brought a list to their non-VA provider appointment, the information was welcomed but fell outside the usual visit workflow. Many provider visits are dominated by current patient symptoms, and issues of reconciling medications may be a lower priority.³² Also, some providers may delegate medication reconciliation functions to a nurse or other support staff. One physician offered that he delegated logging in to a patient’s online medication information to a health coach on staff. These findings were consistent with perspectives shared by non-VA family practice physicians about personal health records.³³

The most common way to integrate outside medication lists into the non-VA provider’s medical record seemed to be scanning the document. Scanning had its limitations because the provider might be unaware of the scanned document, and there were no mechanisms to import medication names and doses. However, the process may improve only the non-VA providers’ records, as they reported that they had no easy or consistent way to transmit medication changes to notes to the VA.

In general, communicating with VA providers was seen as not feasible and not worth their time or effort. It may be beneficial to address this non-VA provider concern because it seems to inhibit the transfer of important health information and the maintenance of a concordant medication record. Information transfer is particularly relevant for veterans who are primarily cared for by non-VA providers and use the VA only to get prescription medications.

In the current approach, non-VA providers have no simple, direct way to update the VA medication list. Transmitting updates carries the risk of inappropriate changes and is concerning if neither or both prescribers consider themselves to be responsible for the patient’s medications. Also, the potential exists for all medication lists to be inaccurate if the lists do not reflect the medications patients take when left on their own. Patient nonadherence rates can exceed 50%, depending on the medication.^34,35 Several interviewed non-VA physicians stressed the importance of asking patients to list the medications they were using during the medication reconciliation process.

This study offers several areas for additional inquiry, including understanding how providers make sense of medication lists from other sources and what technologies can be applied to increase list consistency without increasing the burden on providers.

Practice Implications

Although patient involvement in medication list sharing has the potential to improve information consistency, health systems, providers, and other stakeholders should be cautious in assuming that other prescribers will work to combat medication list entropy, especially if no systems exist to seamlessly incorporate this information into clinic workflow. Devising standardized procedures when patients bring in their records from other providers increases the likelihood that this information will be incorporated into clinical decision making and maintaining up-to-date medication information for patients who use multiple prescribers.

Limitations

These analyses are based on a small sample size (n = 50 for chart review) and (n = 8 for the semistructured interviews) from a single Midwestern state. These findings should be used as evidence for further inquiry.

Conclusion

This study illuminates the level of discrepancies between the medication lists of veteran dual users, including high rates of discrepancies for high-risk medications, such as anticoagulants and opiates. This study also provides evidence of deficiencies in the health care system to decrease medication list entropy that may place veterans at an elevated risk for adverse medication events.

In the U.S., 4.5 million ambulatory care visits occur annually due to adverse drug reactions (ADRs) of prescription medications.¹ Many ADRs are severe, and they result in more than 100,000 death per year.² A significant number of these ADRs are preventable and are a result of inappropriate prescribing.³ It is well-documented that inappropriate prescribing is exacerbated by the number of patients who see multiple prescribers and use many different prescription medications.⁴ This situation results in many versions of a patient’s medication list and in discrepancies across data sources.⁵

Medication list discrepancies have been researched in the context of care transitions between the hospital and home.^6,7 However, less attention has been given to community-dwelling adults who use multiple outpatient prescribers, a practice common among older adults with chronic conditions who see a primary care provider and several specialists.⁴ Also, veterans are a growing patient population who use providers from multiple health care systems.⁸ Up to 70% of veterans enrolled in VA health care use both VA and non-VA providers. These patients are referred to as dual users.^9,10

There has been an increasing push for patients to be more actively engaged in their own health care, including maintenance of their medication list and other personal health information.^11-13 Providers have realized that patients have important experiences and preferences to share about how they use medications at home.^14,15 Research suggests that patient interest and ability to use patient portals is variable and dependent on age, technical abilities, health literacy, and endorsement by their providers.¹⁶ Greater patient engagement in the medication management process is potentially advantageous, especially because providers from different health care systems often lack the capability to share medication list information.^12,17

My HealtheVet, the VA’s patient portal, offers veterans several features. For example, users can securely message providers, refill prescriptions, check appointments, self-enter information, and download their VA health record (including medication history) using the Blue Button (BB) feature. The BB is managed by the HHS to provide consistency across electronic health record platforms.^18,19

This BB medication list gives VA patients the tool they need to inform their providers about the medications they take, particularly dual users. VA patients that use multiple prescribers are subject to medication list discrepancies because of the fragmentation of information.^4,20

Objectives

The objectives of this study were to (1) describe discrepancies between VA medication lists and non-VA provider medication lists for a group of veteran dual users; (2) identify therapeutic duplications in these lists; and (3) contextualize discrepancies by interviewing non-VA providers about their medication reconciliation processes and management of dual use patients.

Methods

This analysis is based on data collected as part of a pilot randomized controlled trial by Turvey and colleagues.²¹ Veterans with a diagnosis of ≥ 1 chronic health condition (eg, diabetes, hypertension) were invited by letter to participate in a study about using online management of their health information. Interested patients were screened to meet additional inclusion criteria, such as taking ≥ 5 medications, receiving care from a non-VA provider, an appointment with a non-VA provider within the study time frame, and access to a computer, online access, and printer. 

Eligible veterans were randomized to receive either (1) BB training (intervention group) instructing patients to download the Continuity of Care Document and bring it to their non-VA provider visit; or (2) a training evaluating medical information online (control group). Training information was mailed, including written materials and phone support, to both groups. The intervention group could also access an online training link.

One of the objectives was to test whether downloading and bringing the health information to a non-VA appointment decreased medication list discrepancies. The sample was small, and differences in discrepancy rates between groups were not significant. Therefore, groups were combined for the present analysis. Visits occurred between December 2013 and December 2014. Greater detail about study design and primary results are available in the study by Turvey and colleagues.²¹

Study procedures were approved by the University of Iowa Institutional Review Board and the Iowa City VA Health Care System Research and Development Committee. All participants provided consent.

Identifying Discrepancies

A 4-phase process was used to address medication discrepancies.^22,23 The first phase defined medication discrepancy categories. The mutually exclusive categories were dose, frequency, and missing discrepancies. In cases where a medication was both dose and frequency discrepant, only dose discrepancy was applied. For missing medications, entities on only the VA list were marked as “non-VA missing” and medications appearing on only the non-VA list would be denoted as “VA missing.” Medications with no discrepancy were marked as such.

Phase 2 involved collecting medication data. Medication lists from the VA medical record were printed at the time of the non-VA provider appointment. Non-VA medication lists were obtained by sending a medical record request for the visit note, medication list, and any associated visit test results to the non-VA provider office within 2 to 3 weeks of the appointment. Patient names from both lists were replaced with unique patient identifiers.

In phase 3, a research assistant abstracted the hard copy medication lists into a database and identified discrepancies. Variables included medication name, dose, frequency, and administration route. Although administration routes were collected, discrepancies were not assessed because this information commonly was not specified. Medications also were coded as prescription or over-the-counter (OTC). Durable medical equipment often was present on VA lists (eg, syringes, test strips) and was excluded from all analyses. Medications also were not coded as discrepant if they were referenced in a visit note as being changed by the non-VA provider. These combined lists were evaluated by the research assistant based on the discrepancy categories specified in phase 1 and were verified by a pharmacist.

Phase 4 involved counting medication discrepancies. Medication discrepancy rates were calculated at the patient level, both descriptively (mean number of discrepancies per patient) and as a proportion of medications discrepant (number of discrepancies divided by total medications).

Identifying Duplications and High-Risk Medications

A pharmacist examined each combined medication list to identify therapeutic duplications, defined as a patient using ≥ 2 medications from the same medication class (eg, patient taking 2 statin drugs) but not 2 drugs for the same condition (eg, fish oil and atorvastatin for dyslipidemia). High-risk medications also were noted, including anticoagulants, certain nonsteroidal anti-inflammatory drugs, oral and injectable hypoglycemics, opioids, sedatives, and hypnotics.^24-26 These medications received special focus because of their link to a high risk for ADRs.²⁷

Descriptive statistics were calculated for patient characteristics and for each discrepancy type, both overall and according to prescription OTC, and high-risk medications. The proportion of discrepant medications was calculated for each category. Bivariate correlations were calculated for select variables to understand potential relationships.

Interviews With Non-VA Providers

All patients were instructed to bring a consent letter and the 1-page questionnaire to their non-VA provider appointment. The questionnaire contained an item asking whether non-VA providers could be contacted for a 15- to 30-minute follow-up interview. The semistructured, qualitative interviews assessed their experiences working with VA providers and VA patients, experiences with VA documents or records, preferences for receiving information from the VA, experience with personal health records, and sharing information with the VA. Eight interviews were conducted, audio-recorded, and transcribed. The goal of the interviews was to explore and understand provider perspectives on managing dual use veterans, including medication reconciliation processes to add context to the interpretation of medication list analysis. Because the data set was relatively small, summaries of each interview were created to highlight main points. These main points were sorted into topics, summarized, and representative quotes were selected.

Results

Fifty veterans were included in the analysis (Table 1). The mean age was 68.5 (SD 6.2); 90% were men. On average, they reported having 6 chronic health conditions and a fair-to-good health status. Based on the combined medication lists from VA and non-VA providers, veterans took an average of 15.8 (SD 7.0) unique medications (combined prescription and OTC/vitamins) and had an average of 10.0 (SD 6.1) all-type discrepancies (Table 2).

Overall, 58% of the prescription medications were discrepant: The most common discrepancy between the 2 lists was medication missing on one of the lists, which occurred 3.9 times per patient on average for prescription medications and 2.8 times per patient for OTCs. Frequency or dose discrepancies also were common between the lists at a rate of 1.9 discrepancies per patient for prescription medications and 1.2 discrepancies per patient for OTCs.

For high-risk medications, opiates and sedative medications had the most discrepancies between the lists because the VA practitioner may not have known that the patient was taking an opiate, although other discrepancies were present (Table 3). Anticoagulant discrepancies were the most consistent, most of these occurring with aspirin. Last, insulin commonly was dose discrepant between the 2 lists, although it also was missing from one list for a number of patients. Overall, high-risk medications shared a discrepancy rate (46.9%) similar to the overall rate.

Twelve therapeutic duplications were identified in the sample.Ten were between-list duplications, that is, “provider A” thought the patient was on a particular medication and “provider B” thought that the patient was on a different medication (Table 4). In 6 instances, within-list duplications were identified (ie, a provider had 2 medications on the list that should not be taken together because they were in the same drug class). In 4 cases, both between- and within-list duplications were present.

Interview Summaries

Veterans and medication. Multiple non-VA providers said that the primary reason veteran patients were going to a VA provider was to obtain discounted medications. The use of the VA for medications also was a way for the non-VA provider to discover that the patient was a veteran. One non-VA provider was particularly concerned about the impact of new or different medications from VA prescribers on efforts to stabilize the patient’s chronic condition.

Several non-VA providers reported that veterans often brought a medication list to the appointment, and several providers recommended the practice to their patients. Non-VA providers preferred to have patients transfer information from VA, sometimes requesting that veterans bring in their records from recent appointments rather than the non-VA provider obtain the information directly from the VA.

Information sharing. Non-VA providers generally preferred hard copies of medication lists and other documents rather than scans because they were more likely to be included in decision making if the documents were presented during the visit. Also, document scans may be buried in the electronic medical record. Some providers mentioned their interest in electronic transfer of medical information like medication lists if the technology were more developed and better integrated.

“I think the long-term vision would be that it should be electronic… it wouldn’t necessarily be feasible at this time. Our system scans paper documents in to an e-version. … but when the patient comes to their encounter 10 days later, you don’t realize the stuff’s there… Having the patient bring them in is probably a more certain way to make sure that it’s actually included in your decision making as a provider.”

Most non-VA providers welcomed more information such as imaging studies because they reported rarely receiving this information from the VA system. Two mentioned the potential for too much information and wanted concise reports should the flow of information increase. Providers had little interest in logging in to a patient’s online health record portal as a delegate for reasons related to complexity, time, privacy, and lack of mechanism to document the information accessed.

Medication reconciliation. Non-VA providers generally reported that patients bringing their own or an outside medication list would prompt a process of medication reconciliation. The providers were interested in making changes to their records based on other lists, but outside data were verified against a patient self-report of actual use before adopting changes.

“I print out my med list of what I have in the computer and then I just check off my list against their list. And then whatever’s remaining, we talk about what the differences are, when they were changed, what they were changed for, if they were taken off of something, and if I don’t agree, then I’ll tell the patient, ‘look, there’s a disagreement here, they’ve told you not to be on this. I want you on this.”

Should a discrepancy arise, non-VA providers generally had a negative view of attempting to contact VA providers. Other mechanisms such as calling a local pharmacy would be done first.

Discussion

This study provided initial evidence that medication list discrepancies exist for dual use veterans. Other studies of medication list discrepancies have linked such inconsistencies to medication-related problems and negative outcomes for patients.²⁷ Although efforts to increase access to care for veterans have advantages related to expediency, consequences to fragmenting care exist. More robust mechanisms for establishing and maintaining medication list consistency are needed, especially given the lack of a universally accepted medical record format or repository. A multifaceted approach, including patient engagement, seems necessary.

This study also showed that discrepancies of high-risk medications are common for veteran participants, placing them at risk for medication-related problems and harm. These risks included dose and frequency discrepancies that could result in over- or underdosing of medications and in medication omissions, which could cause duplicative therapies and unnecessary risks. For example, aspirin frequently was listed on non-VA lists but was omitted from VA medication lists. This could be problematic for patients who present to the VA for a procedure in which no information about aspirin could jeopardize their safety. Insulin doses also were commonly discrepant, which could impact glycemic control.

Many providers also had incomplete prescribing information for opiates. Those prescriptions are particularly relevant given the link between veterans, posttraumatic stress disorder, depression, and substance abuse.^28-30 However, it was beyond the scope of this pilot study to link these discrepancies to ADRs, such as emergency department visits or hospital admissions. Other studies have demonstrated that discrepancies at hospital discharge can result in these types of negative outcomes.^27,31 Subsequent research should determine the clinical significance of discrepancies that occur when veterans are dual users.

The qualitative interviews provided some initial context on prescriber perspectives about the role of veterans participating in the medication list sharing process and personal health records. It seemed that for the portion of patients who brought a list to their non-VA provider appointment, the information was welcomed but fell outside the usual visit workflow. Many provider visits are dominated by current patient symptoms, and issues of reconciling medications may be a lower priority.³² Also, some providers may delegate medication reconciliation functions to a nurse or other support staff. One physician offered that he delegated logging in to a patient’s online medication information to a health coach on staff. These findings were consistent with perspectives shared by non-VA family practice physicians about personal health records.³³

The most common way to integrate outside medication lists into the non-VA provider’s medical record seemed to be scanning the document. Scanning had its limitations because the provider might be unaware of the scanned document, and there were no mechanisms to import medication names and doses. However, the process may improve only the non-VA providers’ records, as they reported that they had no easy or consistent way to transmit medication changes to notes to the VA.

In general, communicating with VA providers was seen as not feasible and not worth their time or effort. It may be beneficial to address this non-VA provider concern because it seems to inhibit the transfer of important health information and the maintenance of a concordant medication record. Information transfer is particularly relevant for veterans who are primarily cared for by non-VA providers and use the VA only to get prescription medications.

In the current approach, non-VA providers have no simple, direct way to update the VA medication list. Transmitting updates carries the risk of inappropriate changes and is concerning if neither or both prescribers consider themselves to be responsible for the patient’s medications. Also, the potential exists for all medication lists to be inaccurate if the lists do not reflect the medications patients take when left on their own. Patient nonadherence rates can exceed 50%, depending on the medication.^34,35 Several interviewed non-VA physicians stressed the importance of asking patients to list the medications they were using during the medication reconciliation process.

This study offers several areas for additional inquiry, including understanding how providers make sense of medication lists from other sources and what technologies can be applied to increase list consistency without increasing the burden on providers.

Practice Implications

Although patient involvement in medication list sharing has the potential to improve information consistency, health systems, providers, and other stakeholders should be cautious in assuming that other prescribers will work to combat medication list entropy, especially if no systems exist to seamlessly incorporate this information into clinic workflow. Devising standardized procedures when patients bring in their records from other providers increases the likelihood that this information will be incorporated into clinical decision making and maintaining up-to-date medication information for patients who use multiple prescribers.

Limitations

These analyses are based on a small sample size (n = 50 for chart review) and (n = 8 for the semistructured interviews) from a single Midwestern state. These findings should be used as evidence for further inquiry.

Conclusion

This study illuminates the level of discrepancies between the medication lists of veteran dual users, including high rates of discrepancies for high-risk medications, such as anticoagulants and opiates. This study also provides evidence of deficiencies in the health care system to decrease medication list entropy that may place veterans at an elevated risk for adverse medication events.

In the U.S., 4.5 million ambulatory care visits occur annually due to adverse drug reactions (ADRs) of prescription medications.¹ Many ADRs are severe, and they result in more than 100,000 death per year.² A significant number of these ADRs are preventable and are a result of inappropriate prescribing.³ It is well-documented that inappropriate prescribing is exacerbated by the number of patients who see multiple prescribers and use many different prescription medications.⁴ This situation results in many versions of a patient’s medication list and in discrepancies across data sources.⁵

Medication list discrepancies have been researched in the context of care transitions between the hospital and home.^6,7 However, less attention has been given to community-dwelling adults who use multiple outpatient prescribers, a practice common among older adults with chronic conditions who see a primary care provider and several specialists.⁴ Also, veterans are a growing patient population who use providers from multiple health care systems.⁸ Up to 70% of veterans enrolled in VA health care use both VA and non-VA providers. These patients are referred to as dual users.^9,10

There has been an increasing push for patients to be more actively engaged in their own health care, including maintenance of their medication list and other personal health information.^11-13 Providers have realized that patients have important experiences and preferences to share about how they use medications at home.^14,15 Research suggests that patient interest and ability to use patient portals is variable and dependent on age, technical abilities, health literacy, and endorsement by their providers.¹⁶ Greater patient engagement in the medication management process is potentially advantageous, especially because providers from different health care systems often lack the capability to share medication list information.^12,17

My HealtheVet, the VA’s patient portal, offers veterans several features. For example, users can securely message providers, refill prescriptions, check appointments, self-enter information, and download their VA health record (including medication history) using the Blue Button (BB) feature. The BB is managed by the HHS to provide consistency across electronic health record platforms.^18,19

This BB medication list gives VA patients the tool they need to inform their providers about the medications they take, particularly dual users. VA patients that use multiple prescribers are subject to medication list discrepancies because of the fragmentation of information.^4,20

Objectives

The objectives of this study were to (1) describe discrepancies between VA medication lists and non-VA provider medication lists for a group of veteran dual users; (2) identify therapeutic duplications in these lists; and (3) contextualize discrepancies by interviewing non-VA providers about their medication reconciliation processes and management of dual use patients.

Methods

This analysis is based on data collected as part of a pilot randomized controlled trial by Turvey and colleagues.²¹ Veterans with a diagnosis of ≥ 1 chronic health condition (eg, diabetes, hypertension) were invited by letter to participate in a study about using online management of their health information. Interested patients were screened to meet additional inclusion criteria, such as taking ≥ 5 medications, receiving care from a non-VA provider, an appointment with a non-VA provider within the study time frame, and access to a computer, online access, and printer. 

Eligible veterans were randomized to receive either (1) BB training (intervention group) instructing patients to download the Continuity of Care Document and bring it to their non-VA provider visit; or (2) a training evaluating medical information online (control group). Training information was mailed, including written materials and phone support, to both groups. The intervention group could also access an online training link.

One of the objectives was to test whether downloading and bringing the health information to a non-VA appointment decreased medication list discrepancies. The sample was small, and differences in discrepancy rates between groups were not significant. Therefore, groups were combined for the present analysis. Visits occurred between December 2013 and December 2014. Greater detail about study design and primary results are available in the study by Turvey and colleagues.²¹

Study procedures were approved by the University of Iowa Institutional Review Board and the Iowa City VA Health Care System Research and Development Committee. All participants provided consent.

Identifying Discrepancies

A 4-phase process was used to address medication discrepancies.^22,23 The first phase defined medication discrepancy categories. The mutually exclusive categories were dose, frequency, and missing discrepancies. In cases where a medication was both dose and frequency discrepant, only dose discrepancy was applied. For missing medications, entities on only the VA list were marked as “non-VA missing” and medications appearing on only the non-VA list would be denoted as “VA missing.” Medications with no discrepancy were marked as such.

Phase 2 involved collecting medication data. Medication lists from the VA medical record were printed at the time of the non-VA provider appointment. Non-VA medication lists were obtained by sending a medical record request for the visit note, medication list, and any associated visit test results to the non-VA provider office within 2 to 3 weeks of the appointment. Patient names from both lists were replaced with unique patient identifiers.

In phase 3, a research assistant abstracted the hard copy medication lists into a database and identified discrepancies. Variables included medication name, dose, frequency, and administration route. Although administration routes were collected, discrepancies were not assessed because this information commonly was not specified. Medications also were coded as prescription or over-the-counter (OTC). Durable medical equipment often was present on VA lists (eg, syringes, test strips) and was excluded from all analyses. Medications also were not coded as discrepant if they were referenced in a visit note as being changed by the non-VA provider. These combined lists were evaluated by the research assistant based on the discrepancy categories specified in phase 1 and were verified by a pharmacist.

Phase 4 involved counting medication discrepancies. Medication discrepancy rates were calculated at the patient level, both descriptively (mean number of discrepancies per patient) and as a proportion of medications discrepant (number of discrepancies divided by total medications).

Identifying Duplications and High-Risk Medications

A pharmacist examined each combined medication list to identify therapeutic duplications, defined as a patient using ≥ 2 medications from the same medication class (eg, patient taking 2 statin drugs) but not 2 drugs for the same condition (eg, fish oil and atorvastatin for dyslipidemia). High-risk medications also were noted, including anticoagulants, certain nonsteroidal anti-inflammatory drugs, oral and injectable hypoglycemics, opioids, sedatives, and hypnotics.^24-26 These medications received special focus because of their link to a high risk for ADRs.²⁷

Descriptive statistics were calculated for patient characteristics and for each discrepancy type, both overall and according to prescription OTC, and high-risk medications. The proportion of discrepant medications was calculated for each category. Bivariate correlations were calculated for select variables to understand potential relationships.

Interviews With Non-VA Providers

All patients were instructed to bring a consent letter and the 1-page questionnaire to their non-VA provider appointment. The questionnaire contained an item asking whether non-VA providers could be contacted for a 15- to 30-minute follow-up interview. The semistructured, qualitative interviews assessed their experiences working with VA providers and VA patients, experiences with VA documents or records, preferences for receiving information from the VA, experience with personal health records, and sharing information with the VA. Eight interviews were conducted, audio-recorded, and transcribed. The goal of the interviews was to explore and understand provider perspectives on managing dual use veterans, including medication reconciliation processes to add context to the interpretation of medication list analysis. Because the data set was relatively small, summaries of each interview were created to highlight main points. These main points were sorted into topics, summarized, and representative quotes were selected.

Results

Fifty veterans were included in the analysis (Table 1). The mean age was 68.5 (SD 6.2); 90% were men. On average, they reported having 6 chronic health conditions and a fair-to-good health status. Based on the combined medication lists from VA and non-VA providers, veterans took an average of 15.8 (SD 7.0) unique medications (combined prescription and OTC/vitamins) and had an average of 10.0 (SD 6.1) all-type discrepancies (Table 2).

Overall, 58% of the prescription medications were discrepant: The most common discrepancy between the 2 lists was medication missing on one of the lists, which occurred 3.9 times per patient on average for prescription medications and 2.8 times per patient for OTCs. Frequency or dose discrepancies also were common between the lists at a rate of 1.9 discrepancies per patient for prescription medications and 1.2 discrepancies per patient for OTCs.

For high-risk medications, opiates and sedative medications had the most discrepancies between the lists because the VA practitioner may not have known that the patient was taking an opiate, although other discrepancies were present (Table 3). Anticoagulant discrepancies were the most consistent, most of these occurring with aspirin. Last, insulin commonly was dose discrepant between the 2 lists, although it also was missing from one list for a number of patients. Overall, high-risk medications shared a discrepancy rate (46.9%) similar to the overall rate.

Twelve therapeutic duplications were identified in the sample.Ten were between-list duplications, that is, “provider A” thought the patient was on a particular medication and “provider B” thought that the patient was on a different medication (Table 4). In 6 instances, within-list duplications were identified (ie, a provider had 2 medications on the list that should not be taken together because they were in the same drug class). In 4 cases, both between- and within-list duplications were present.

Interview Summaries

Veterans and medication. Multiple non-VA providers said that the primary reason veteran patients were going to a VA provider was to obtain discounted medications. The use of the VA for medications also was a way for the non-VA provider to discover that the patient was a veteran. One non-VA provider was particularly concerned about the impact of new or different medications from VA prescribers on efforts to stabilize the patient’s chronic condition.

Several non-VA providers reported that veterans often brought a medication list to the appointment, and several providers recommended the practice to their patients. Non-VA providers preferred to have patients transfer information from VA, sometimes requesting that veterans bring in their records from recent appointments rather than the non-VA provider obtain the information directly from the VA.

Information sharing. Non-VA providers generally preferred hard copies of medication lists and other documents rather than scans because they were more likely to be included in decision making if the documents were presented during the visit. Also, document scans may be buried in the electronic medical record. Some providers mentioned their interest in electronic transfer of medical information like medication lists if the technology were more developed and better integrated.

“I think the long-term vision would be that it should be electronic… it wouldn’t necessarily be feasible at this time. Our system scans paper documents in to an e-version. … but when the patient comes to their encounter 10 days later, you don’t realize the stuff’s there… Having the patient bring them in is probably a more certain way to make sure that it’s actually included in your decision making as a provider.”

Most non-VA providers welcomed more information such as imaging studies because they reported rarely receiving this information from the VA system. Two mentioned the potential for too much information and wanted concise reports should the flow of information increase. Providers had little interest in logging in to a patient’s online health record portal as a delegate for reasons related to complexity, time, privacy, and lack of mechanism to document the information accessed.

Medication reconciliation. Non-VA providers generally reported that patients bringing their own or an outside medication list would prompt a process of medication reconciliation. The providers were interested in making changes to their records based on other lists, but outside data were verified against a patient self-report of actual use before adopting changes.

“I print out my med list of what I have in the computer and then I just check off my list against their list. And then whatever’s remaining, we talk about what the differences are, when they were changed, what they were changed for, if they were taken off of something, and if I don’t agree, then I’ll tell the patient, ‘look, there’s a disagreement here, they’ve told you not to be on this. I want you on this.”

Should a discrepancy arise, non-VA providers generally had a negative view of attempting to contact VA providers. Other mechanisms such as calling a local pharmacy would be done first.

Discussion

This study provided initial evidence that medication list discrepancies exist for dual use veterans. Other studies of medication list discrepancies have linked such inconsistencies to medication-related problems and negative outcomes for patients.²⁷ Although efforts to increase access to care for veterans have advantages related to expediency, consequences to fragmenting care exist. More robust mechanisms for establishing and maintaining medication list consistency are needed, especially given the lack of a universally accepted medical record format or repository. A multifaceted approach, including patient engagement, seems necessary.

This study also showed that discrepancies of high-risk medications are common for veteran participants, placing them at risk for medication-related problems and harm. These risks included dose and frequency discrepancies that could result in over- or underdosing of medications and in medication omissions, which could cause duplicative therapies and unnecessary risks. For example, aspirin frequently was listed on non-VA lists but was omitted from VA medication lists. This could be problematic for patients who present to the VA for a procedure in which no information about aspirin could jeopardize their safety. Insulin doses also were commonly discrepant, which could impact glycemic control.

Many providers also had incomplete prescribing information for opiates. Those prescriptions are particularly relevant given the link between veterans, posttraumatic stress disorder, depression, and substance abuse.^28-30 However, it was beyond the scope of this pilot study to link these discrepancies to ADRs, such as emergency department visits or hospital admissions. Other studies have demonstrated that discrepancies at hospital discharge can result in these types of negative outcomes.^27,31 Subsequent research should determine the clinical significance of discrepancies that occur when veterans are dual users.

The qualitative interviews provided some initial context on prescriber perspectives about the role of veterans participating in the medication list sharing process and personal health records. It seemed that for the portion of patients who brought a list to their non-VA provider appointment, the information was welcomed but fell outside the usual visit workflow. Many provider visits are dominated by current patient symptoms, and issues of reconciling medications may be a lower priority.³² Also, some providers may delegate medication reconciliation functions to a nurse or other support staff. One physician offered that he delegated logging in to a patient’s online medication information to a health coach on staff. These findings were consistent with perspectives shared by non-VA family practice physicians about personal health records.³³

The most common way to integrate outside medication lists into the non-VA provider’s medical record seemed to be scanning the document. Scanning had its limitations because the provider might be unaware of the scanned document, and there were no mechanisms to import medication names and doses. However, the process may improve only the non-VA providers’ records, as they reported that they had no easy or consistent way to transmit medication changes to notes to the VA.

In general, communicating with VA providers was seen as not feasible and not worth their time or effort. It may be beneficial to address this non-VA provider concern because it seems to inhibit the transfer of important health information and the maintenance of a concordant medication record. Information transfer is particularly relevant for veterans who are primarily cared for by non-VA providers and use the VA only to get prescription medications.

In the current approach, non-VA providers have no simple, direct way to update the VA medication list. Transmitting updates carries the risk of inappropriate changes and is concerning if neither or both prescribers consider themselves to be responsible for the patient’s medications. Also, the potential exists for all medication lists to be inaccurate if the lists do not reflect the medications patients take when left on their own. Patient nonadherence rates can exceed 50%, depending on the medication.^34,35 Several interviewed non-VA physicians stressed the importance of asking patients to list the medications they were using during the medication reconciliation process.

This study offers several areas for additional inquiry, including understanding how providers make sense of medication lists from other sources and what technologies can be applied to increase list consistency without increasing the burden on providers.

Practice Implications

Although patient involvement in medication list sharing has the potential to improve information consistency, health systems, providers, and other stakeholders should be cautious in assuming that other prescribers will work to combat medication list entropy, especially if no systems exist to seamlessly incorporate this information into clinic workflow. Devising standardized procedures when patients bring in their records from other providers increases the likelihood that this information will be incorporated into clinical decision making and maintaining up-to-date medication information for patients who use multiple prescribers.

Limitations

These analyses are based on a small sample size (n = 50 for chart review) and (n = 8 for the semistructured interviews) from a single Midwestern state. These findings should be used as evidence for further inquiry.

Conclusion

This study illuminates the level of discrepancies between the medication lists of veteran dual users, including high rates of discrepancies for high-risk medications, such as anticoagulants and opiates. This study also provides evidence of deficiencies in the health care system to decrease medication list entropy that may place veterans at an elevated risk for adverse medication events.

Acne vulgaris predominantly occurs during puberty and can persist beyond 25 years of age, most commonly in women.^1,2 Although acne does not cause physical impairment, it can be associated with a considerable psychosocial burden including increased levels of anxiety, anger, depression, and frustration, which in turn can affect vocational and academic performance, quality of life (QOL), and self-esteem.³

Quality of life measures provide valuable insight into the debilitating effects of acne.¹ It has been suggested that acne patients may experience poor body image and low self-esteem as well as social isolation and constriction of activities.⁴ Self-esteem is a favorable and unfavorable attitude toward oneself.⁵ A marked emphasis has been placed on body image in society, fueled by external cues such as the media.^3,6 This study was carried out to assess QOL and self-esteem in acne patients.

Methods

This prospective, hospital-based, cross-sectional, case-control study was conducted at The Oxford Medical College, Hospital & Research Center (Bangalore, India), over a period of 3 months. One hundred consecutive acne cases (age range, 12–45 years) and 100 age- and gender-matched controls who did not have any skin disease provided consent and were included in the analysis. Guardians gave consent for individuals who were younger than 18 years. Exclusion criteria for cases included a medical disorder (eg, epilepsy, diabetes mellitus, hypertension) or medications that would likely interfere with acne assessment.

The cases and controls were administered a semistructured questionnaire to collect sociodemographic details. Acne was graded for the predominant lesions, QOL was assessed using the Cardiff Acne Disability Index (CADI) and World Health Organization Quality of Life–BREF (WHOQOL-BREF) scale, and self-esteem was measured using the Rosenberg self-esteem scale (RSES). The study was approved by the institutional review board.

Acne Grading

Acne was graded according to the predominant lesions using the following criteria: grade 1=comedones and occasional papules; grade 2=papules, comedones, and few pustules; grade 3=predominant pustules, nodules, and abscesses; and grade 4=mainly cysts, abscesses, and widespread scarring.¹

Quality of Life Assessment

The CADI questionnaire was used to assess the level of disability caused by acne.⁶ It is a 5-item questionnaire with scores ranging from 0 to 3 for a total maximum score of 15 and minimum score of 0. Total scores were classified as low (0–4), medium (5–9), and high (10–15).⁷

The WHOQOL-BREF is a self-reported questionnaire containing 26 items that make up the 4 domains of physical health (7 items), psychological health (6 items), social relationships (3 items), and environment (8 items); there also are 2 single questions regarding the overall perception of QOL and health. Questions were scored on aseries of 5-point scales with higher scores denoting better QOL.⁸

Self-esteem Assessment

The RSES uses a 5-point Likert scale from strongly agree to strongly disagree to rate a series of 10 statements. The total score ranges from 0 to 30. Scores less than 15 suggest low self-esteem, while scores of 15 and greater indicate high self-esteem.⁵

Statistical Analysis

Results were analyzed using descriptive and inferential statistical methods. A χ² test was used for categorical data, and a Student t test and an analysis of variance were used for continuous data.

Results

The study consisted of 100 cases and 100 controls. The mean age was 21 years. The majority of cases reported an age of onset of acne of 11 to 20 years (66%), were predominantly female (58%) from rural backgrounds, and had a family history of acne (68%). The majority of lesions ceased within 24 months (60%). The face was the most commonly involved area (80%) and papules were the most prevalent lesion type (62%).

Cases predominantly had grade 2 acne (46%), and there was medium to high impairment in QOL according to CADI scores.

The scores for all the domains of the WHOQOL-BREF as well as the total score were lower in cases compared to controls (Table). There was a statistically significant difference between the 2 groups in the psychological (P=.0402) and environment (P=.006) domains.

The RSES mean (SD) score was higher in controls (19.74 [4.23]) than in cases (15.72 [5.06]) and was statistically significant (P<.0001). Low self-esteem was noted in 38% of cases and 16% of controls, and high self-esteem was noted in 62% and 84%, respectively.

In reviewing the correlation between acne severity, CADI, WHOQOL-BREF, and RSES scores, we found a positive correlation between acne severity and CADI scores (R=0.51), which implies that as the severity of acne worsens, the QOL impairment increases. There was a negative correlation between acne severity, WHOQOL-BREF score (R=–0.13), and RSES score (R=–0.18), which showed that as the severity of acne increases, QOL and self-esteem decrease. We observed that as the grade of acne increases, there is a statistically significant impairment in the QOL according to CADI (P<.001), while there is a reduction in QOL and self-esteem according to WHOQOL-BREF and RSES, respectively (P>.05).

Comment

Patients are more likely to develop acne than any other skin disease in their lifetime. Only in recent years has the psychodermatologic literature begun to address the possibility of acne having a psychological and emotional impact.⁴ Although the cause-and-effect relationship between acne and psychological trauma has been debated for decades, only recently has the measurement focus shifted from psychological correlates (eg, personality) and emotional triggers (eg, stress) to the effect of acne on patients’ QOL and self-esteem. This shift occurred as validated instruments for measuring disability, QOL, and self-esteem, specifically in patients with skin diseases, became available.⁹

In our study, the age of onset of acne was 11 to 20 years and it affected predominantly females (58%), which is in concordance with other studies, as acne develops in adolescence and subsides in adulthood.^1,10 Acne is more common in females due to hormonal factors and use of cosmetics. We observed that the face (80%) was most frequently affected, followed by the back (14%) and chest (6%), which is similar to prior studies.^1,10 Because the face plays an important role in body image, the presence of facial lesions may be unacceptable for patients and therefore they may present more frequently to dermatologists.

In our study, 68% of cases and 22% of controls had a family history of acne. A similar correlation also was noted in other studies, which suggests acne has an inherited predisposition due to involvement of the cytochrome P450-1A1 gene, CYP1A1, and steroid 21-hydroxylase, P-450-c21.^1,11 We found 46% of cases had grade 2 acne and 36% had grade 1 acne, which was congruent with prior studies.^12,13 Patients with severe acne are more likely to seek medical intervention in hospitals.

In our study, 58% of the cases had medium to high impairment in QOL according to CADI scores. We noticed as the severity of acne increased there was severe impairment in QOL. Similar findings have been found in studies that used other scales to assess QOL.^1,6,9

In our study, 38% of cases and 16% of controls had low self-esteem, which was statistically significant (P<.0001). There was a negative correlation between the severity of acne and self-esteem. In a prior study of 240 professional college students, 53% had feelings of low self-esteem and 40% revealed they avoided social gatherings and interactions with the opposite sex because of their acne.¹⁴ In a questionnaire-based survey of 3775 students, it was observed that the presence of acne correlated with poor self-attitude in boys and poor self-worth in girls.³ We found patients with grade 1 acne had higher self-esteem as compared to other grades of acne. Similarly, a cross-sectional study by Uslu et al¹⁵ found a direct correlation between acne severity and lower self-esteem using the RSES questionnaire. Although acne may be viewed as a minor cosmetic issue, it can have a negative impact on self-esteem and interpersonal relationships. Many of the studies had not used a validated structured questionnaire to assess self-esteem and there is a paucity of literature in relation to acne and self-esteem.^3,16,17

According to the WHOQOL-BREF, the psychological domain was affected more in cases than in controls, which was a statistically significant difference. One study observed that patients experience immediate psychological consequences of acne such as reduced self-esteem, poor self-image, self-consciousness, and embarrassment.³ These effects are exacerbated by taunting, stigmatization, and perceptions of scrutiny and being judged, causing patients to avoid interaction and social situations. Similarly, Pruthi and Babu¹⁸ observed that acne had an impact on the psychosocial aspects of adult females using the Dermatology Life Quality Index and CADI.

Financial resources, health and social care accessibility, and opportunities for acquiring new information and skills were the factors that were considered in the environment domain of the WHOQOL-BREF.⁸ We noted that the environment domain scores were significantly lower in cases than in controls. The cases could have had a detrimental effect on the latest opportunities in occupational functioning due to acne, and as most of the population was from a rural area, they were having less favorable circumstances in acquiring new information about the management of acne.

There was no statistically significant difference between cases and controls in the social and physical domains of the WHOQOL-BREF, which suggests that these fields do not influence QOL. Similarly, patients in Sarawak, Malaysia, were least affected in the domain of social functioning, which was likely attributed to the upbringing of this population encouraging stoicism.¹⁹

In the current study, QOL impairment showed a positive correlation with acne severity according to CADI scores; however, there was no significant difference between WHOQOL-BREF score and acne grading, which suggests that QOL impairment does not depend on severity of acne alone. Physical, psychological, social, and environment domains play an important role in impaired QOL. Hence, by using the WHOQOL-BREF we can evaluate the actual domain that is adversely affected by acne and can be treated with a holistic approach. This point must be stressed in the training of medical faculty, as the treatment of acne should not be based on acne severity alone but also on the degree of QOL impairment.¹⁹

These results indicate that more data are required and there is a need to consider other variables that could play a role. This study was a hospital-based, cross-sectional study with a small sample group that cannot be generalized, which are limitations. Longitudinal follow-up of the cases before and after treatment was not done. The questionnaires helped us to detect psychosocial aspects but were insufficient to diagnose psychiatric comorbidity.

The strengths of this study include the use of a specific scale for the assessment of self-esteem. The usage of comprehensive (WHOQOL-BREF) and specific (CADI) scales to evaluate QOL has mutual advantage.

Conclusion

Acne vulgaris is a disease that can adversely affect an individual’s QOL and self-esteem. This study suggested the importance of screening for psychosocial problems in those who present for management of acne. It is important for dermatologists to be cautious about psychological problems in acne patients and be aware of the importance of basic psychosomatic treatment in conjunction with medical treatment in the management of acne.

Acne vulgaris predominantly occurs during puberty and can persist beyond 25 years of age, most commonly in women.^1,2 Although acne does not cause physical impairment, it can be associated with a considerable psychosocial burden including increased levels of anxiety, anger, depression, and frustration, which in turn can affect vocational and academic performance, quality of life (QOL), and self-esteem.³

Quality of life measures provide valuable insight into the debilitating effects of acne.¹ It has been suggested that acne patients may experience poor body image and low self-esteem as well as social isolation and constriction of activities.⁴ Self-esteem is a favorable and unfavorable attitude toward oneself.⁵ A marked emphasis has been placed on body image in society, fueled by external cues such as the media.^3,6 This study was carried out to assess QOL and self-esteem in acne patients.

Methods

This prospective, hospital-based, cross-sectional, case-control study was conducted at The Oxford Medical College, Hospital & Research Center (Bangalore, India), over a period of 3 months. One hundred consecutive acne cases (age range, 12–45 years) and 100 age- and gender-matched controls who did not have any skin disease provided consent and were included in the analysis. Guardians gave consent for individuals who were younger than 18 years. Exclusion criteria for cases included a medical disorder (eg, epilepsy, diabetes mellitus, hypertension) or medications that would likely interfere with acne assessment.

The cases and controls were administered a semistructured questionnaire to collect sociodemographic details. Acne was graded for the predominant lesions, QOL was assessed using the Cardiff Acne Disability Index (CADI) and World Health Organization Quality of Life–BREF (WHOQOL-BREF) scale, and self-esteem was measured using the Rosenberg self-esteem scale (RSES). The study was approved by the institutional review board.

Acne Grading

Acne was graded according to the predominant lesions using the following criteria: grade 1=comedones and occasional papules; grade 2=papules, comedones, and few pustules; grade 3=predominant pustules, nodules, and abscesses; and grade 4=mainly cysts, abscesses, and widespread scarring.¹

Quality of Life Assessment

The CADI questionnaire was used to assess the level of disability caused by acne.⁶ It is a 5-item questionnaire with scores ranging from 0 to 3 for a total maximum score of 15 and minimum score of 0. Total scores were classified as low (0–4), medium (5–9), and high (10–15).⁷

The WHOQOL-BREF is a self-reported questionnaire containing 26 items that make up the 4 domains of physical health (7 items), psychological health (6 items), social relationships (3 items), and environment (8 items); there also are 2 single questions regarding the overall perception of QOL and health. Questions were scored on aseries of 5-point scales with higher scores denoting better QOL.⁸

Self-esteem Assessment

The RSES uses a 5-point Likert scale from strongly agree to strongly disagree to rate a series of 10 statements. The total score ranges from 0 to 30. Scores less than 15 suggest low self-esteem, while scores of 15 and greater indicate high self-esteem.⁵

Statistical Analysis

Results were analyzed using descriptive and inferential statistical methods. A χ² test was used for categorical data, and a Student t test and an analysis of variance were used for continuous data.

Results

The study consisted of 100 cases and 100 controls. The mean age was 21 years. The majority of cases reported an age of onset of acne of 11 to 20 years (66%), were predominantly female (58%) from rural backgrounds, and had a family history of acne (68%). The majority of lesions ceased within 24 months (60%). The face was the most commonly involved area (80%) and papules were the most prevalent lesion type (62%).

Cases predominantly had grade 2 acne (46%), and there was medium to high impairment in QOL according to CADI scores.

The scores for all the domains of the WHOQOL-BREF as well as the total score were lower in cases compared to controls (Table). There was a statistically significant difference between the 2 groups in the psychological (P=.0402) and environment (P=.006) domains.

The RSES mean (SD) score was higher in controls (19.74 [4.23]) than in cases (15.72 [5.06]) and was statistically significant (P<.0001). Low self-esteem was noted in 38% of cases and 16% of controls, and high self-esteem was noted in 62% and 84%, respectively.

In reviewing the correlation between acne severity, CADI, WHOQOL-BREF, and RSES scores, we found a positive correlation between acne severity and CADI scores (R=0.51), which implies that as the severity of acne worsens, the QOL impairment increases. There was a negative correlation between acne severity, WHOQOL-BREF score (R=–0.13), and RSES score (R=–0.18), which showed that as the severity of acne increases, QOL and self-esteem decrease. We observed that as the grade of acne increases, there is a statistically significant impairment in the QOL according to CADI (P<.001), while there is a reduction in QOL and self-esteem according to WHOQOL-BREF and RSES, respectively (P>.05).

Comment

Patients are more likely to develop acne than any other skin disease in their lifetime. Only in recent years has the psychodermatologic literature begun to address the possibility of acne having a psychological and emotional impact.⁴ Although the cause-and-effect relationship between acne and psychological trauma has been debated for decades, only recently has the measurement focus shifted from psychological correlates (eg, personality) and emotional triggers (eg, stress) to the effect of acne on patients’ QOL and self-esteem. This shift occurred as validated instruments for measuring disability, QOL, and self-esteem, specifically in patients with skin diseases, became available.⁹

In our study, the age of onset of acne was 11 to 20 years and it affected predominantly females (58%), which is in concordance with other studies, as acne develops in adolescence and subsides in adulthood.^1,10 Acne is more common in females due to hormonal factors and use of cosmetics. We observed that the face (80%) was most frequently affected, followed by the back (14%) and chest (6%), which is similar to prior studies.^1,10 Because the face plays an important role in body image, the presence of facial lesions may be unacceptable for patients and therefore they may present more frequently to dermatologists.

In our study, 68% of cases and 22% of controls had a family history of acne. A similar correlation also was noted in other studies, which suggests acne has an inherited predisposition due to involvement of the cytochrome P450-1A1 gene, CYP1A1, and steroid 21-hydroxylase, P-450-c21.^1,11 We found 46% of cases had grade 2 acne and 36% had grade 1 acne, which was congruent with prior studies.^12,13 Patients with severe acne are more likely to seek medical intervention in hospitals.

In our study, 58% of the cases had medium to high impairment in QOL according to CADI scores. We noticed as the severity of acne increased there was severe impairment in QOL. Similar findings have been found in studies that used other scales to assess QOL.^1,6,9

In our study, 38% of cases and 16% of controls had low self-esteem, which was statistically significant (P<.0001). There was a negative correlation between the severity of acne and self-esteem. In a prior study of 240 professional college students, 53% had feelings of low self-esteem and 40% revealed they avoided social gatherings and interactions with the opposite sex because of their acne.¹⁴ In a questionnaire-based survey of 3775 students, it was observed that the presence of acne correlated with poor self-attitude in boys and poor self-worth in girls.³ We found patients with grade 1 acne had higher self-esteem as compared to other grades of acne. Similarly, a cross-sectional study by Uslu et al¹⁵ found a direct correlation between acne severity and lower self-esteem using the RSES questionnaire. Although acne may be viewed as a minor cosmetic issue, it can have a negative impact on self-esteem and interpersonal relationships. Many of the studies had not used a validated structured questionnaire to assess self-esteem and there is a paucity of literature in relation to acne and self-esteem.^3,16,17

According to the WHOQOL-BREF, the psychological domain was affected more in cases than in controls, which was a statistically significant difference. One study observed that patients experience immediate psychological consequences of acne such as reduced self-esteem, poor self-image, self-consciousness, and embarrassment.³ These effects are exacerbated by taunting, stigmatization, and perceptions of scrutiny and being judged, causing patients to avoid interaction and social situations. Similarly, Pruthi and Babu¹⁸ observed that acne had an impact on the psychosocial aspects of adult females using the Dermatology Life Quality Index and CADI.

Financial resources, health and social care accessibility, and opportunities for acquiring new information and skills were the factors that were considered in the environment domain of the WHOQOL-BREF.⁸ We noted that the environment domain scores were significantly lower in cases than in controls. The cases could have had a detrimental effect on the latest opportunities in occupational functioning due to acne, and as most of the population was from a rural area, they were having less favorable circumstances in acquiring new information about the management of acne.

There was no statistically significant difference between cases and controls in the social and physical domains of the WHOQOL-BREF, which suggests that these fields do not influence QOL. Similarly, patients in Sarawak, Malaysia, were least affected in the domain of social functioning, which was likely attributed to the upbringing of this population encouraging stoicism.¹⁹

In the current study, QOL impairment showed a positive correlation with acne severity according to CADI scores; however, there was no significant difference between WHOQOL-BREF score and acne grading, which suggests that QOL impairment does not depend on severity of acne alone. Physical, psychological, social, and environment domains play an important role in impaired QOL. Hence, by using the WHOQOL-BREF we can evaluate the actual domain that is adversely affected by acne and can be treated with a holistic approach. This point must be stressed in the training of medical faculty, as the treatment of acne should not be based on acne severity alone but also on the degree of QOL impairment.¹⁹

These results indicate that more data are required and there is a need to consider other variables that could play a role. This study was a hospital-based, cross-sectional study with a small sample group that cannot be generalized, which are limitations. Longitudinal follow-up of the cases before and after treatment was not done. The questionnaires helped us to detect psychosocial aspects but were insufficient to diagnose psychiatric comorbidity.

The strengths of this study include the use of a specific scale for the assessment of self-esteem. The usage of comprehensive (WHOQOL-BREF) and specific (CADI) scales to evaluate QOL has mutual advantage.

Conclusion

Acne vulgaris is a disease that can adversely affect an individual’s QOL and self-esteem. This study suggested the importance of screening for psychosocial problems in those who present for management of acne. It is important for dermatologists to be cautious about psychological problems in acne patients and be aware of the importance of basic psychosomatic treatment in conjunction with medical treatment in the management of acne.

Acne vulgaris predominantly occurs during puberty and can persist beyond 25 years of age, most commonly in women.^1,2 Although acne does not cause physical impairment, it can be associated with a considerable psychosocial burden including increased levels of anxiety, anger, depression, and frustration, which in turn can affect vocational and academic performance, quality of life (QOL), and self-esteem.³

Quality of life measures provide valuable insight into the debilitating effects of acne.¹ It has been suggested that acne patients may experience poor body image and low self-esteem as well as social isolation and constriction of activities.⁴ Self-esteem is a favorable and unfavorable attitude toward oneself.⁵ A marked emphasis has been placed on body image in society, fueled by external cues such as the media.^3,6 This study was carried out to assess QOL and self-esteem in acne patients.

Methods

This prospective, hospital-based, cross-sectional, case-control study was conducted at The Oxford Medical College, Hospital & Research Center (Bangalore, India), over a period of 3 months. One hundred consecutive acne cases (age range, 12–45 years) and 100 age- and gender-matched controls who did not have any skin disease provided consent and were included in the analysis. Guardians gave consent for individuals who were younger than 18 years. Exclusion criteria for cases included a medical disorder (eg, epilepsy, diabetes mellitus, hypertension) or medications that would likely interfere with acne assessment.

The cases and controls were administered a semistructured questionnaire to collect sociodemographic details. Acne was graded for the predominant lesions, QOL was assessed using the Cardiff Acne Disability Index (CADI) and World Health Organization Quality of Life–BREF (WHOQOL-BREF) scale, and self-esteem was measured using the Rosenberg self-esteem scale (RSES). The study was approved by the institutional review board.

Acne Grading

Acne was graded according to the predominant lesions using the following criteria: grade 1=comedones and occasional papules; grade 2=papules, comedones, and few pustules; grade 3=predominant pustules, nodules, and abscesses; and grade 4=mainly cysts, abscesses, and widespread scarring.¹

Quality of Life Assessment

The CADI questionnaire was used to assess the level of disability caused by acne.⁶ It is a 5-item questionnaire with scores ranging from 0 to 3 for a total maximum score of 15 and minimum score of 0. Total scores were classified as low (0–4), medium (5–9), and high (10–15).⁷

The WHOQOL-BREF is a self-reported questionnaire containing 26 items that make up the 4 domains of physical health (7 items), psychological health (6 items), social relationships (3 items), and environment (8 items); there also are 2 single questions regarding the overall perception of QOL and health. Questions were scored on aseries of 5-point scales with higher scores denoting better QOL.⁸

Self-esteem Assessment

The RSES uses a 5-point Likert scale from strongly agree to strongly disagree to rate a series of 10 statements. The total score ranges from 0 to 30. Scores less than 15 suggest low self-esteem, while scores of 15 and greater indicate high self-esteem.⁵

Statistical Analysis

Results were analyzed using descriptive and inferential statistical methods. A χ² test was used for categorical data, and a Student t test and an analysis of variance were used for continuous data.

Results

The study consisted of 100 cases and 100 controls. The mean age was 21 years. The majority of cases reported an age of onset of acne of 11 to 20 years (66%), were predominantly female (58%) from rural backgrounds, and had a family history of acne (68%). The majority of lesions ceased within 24 months (60%). The face was the most commonly involved area (80%) and papules were the most prevalent lesion type (62%).

Cases predominantly had grade 2 acne (46%), and there was medium to high impairment in QOL according to CADI scores.

The scores for all the domains of the WHOQOL-BREF as well as the total score were lower in cases compared to controls (Table). There was a statistically significant difference between the 2 groups in the psychological (P=.0402) and environment (P=.006) domains.

The RSES mean (SD) score was higher in controls (19.74 [4.23]) than in cases (15.72 [5.06]) and was statistically significant (P<.0001). Low self-esteem was noted in 38% of cases and 16% of controls, and high self-esteem was noted in 62% and 84%, respectively.

In reviewing the correlation between acne severity, CADI, WHOQOL-BREF, and RSES scores, we found a positive correlation between acne severity and CADI scores (R=0.51), which implies that as the severity of acne worsens, the QOL impairment increases. There was a negative correlation between acne severity, WHOQOL-BREF score (R=–0.13), and RSES score (R=–0.18), which showed that as the severity of acne increases, QOL and self-esteem decrease. We observed that as the grade of acne increases, there is a statistically significant impairment in the QOL according to CADI (P<.001), while there is a reduction in QOL and self-esteem according to WHOQOL-BREF and RSES, respectively (P>.05).

Comment

Patients are more likely to develop acne than any other skin disease in their lifetime. Only in recent years has the psychodermatologic literature begun to address the possibility of acne having a psychological and emotional impact.⁴ Although the cause-and-effect relationship between acne and psychological trauma has been debated for decades, only recently has the measurement focus shifted from psychological correlates (eg, personality) and emotional triggers (eg, stress) to the effect of acne on patients’ QOL and self-esteem. This shift occurred as validated instruments for measuring disability, QOL, and self-esteem, specifically in patients with skin diseases, became available.⁹

In our study, the age of onset of acne was 11 to 20 years and it affected predominantly females (58%), which is in concordance with other studies, as acne develops in adolescence and subsides in adulthood.^1,10 Acne is more common in females due to hormonal factors and use of cosmetics. We observed that the face (80%) was most frequently affected, followed by the back (14%) and chest (6%), which is similar to prior studies.^1,10 Because the face plays an important role in body image, the presence of facial lesions may be unacceptable for patients and therefore they may present more frequently to dermatologists.

In our study, 68% of cases and 22% of controls had a family history of acne. A similar correlation also was noted in other studies, which suggests acne has an inherited predisposition due to involvement of the cytochrome P450-1A1 gene, CYP1A1, and steroid 21-hydroxylase, P-450-c21.^1,11 We found 46% of cases had grade 2 acne and 36% had grade 1 acne, which was congruent with prior studies.^12,13 Patients with severe acne are more likely to seek medical intervention in hospitals.

In our study, 58% of the cases had medium to high impairment in QOL according to CADI scores. We noticed as the severity of acne increased there was severe impairment in QOL. Similar findings have been found in studies that used other scales to assess QOL.^1,6,9

In our study, 38% of cases and 16% of controls had low self-esteem, which was statistically significant (P<.0001). There was a negative correlation between the severity of acne and self-esteem. In a prior study of 240 professional college students, 53% had feelings of low self-esteem and 40% revealed they avoided social gatherings and interactions with the opposite sex because of their acne.¹⁴ In a questionnaire-based survey of 3775 students, it was observed that the presence of acne correlated with poor self-attitude in boys and poor self-worth in girls.³ We found patients with grade 1 acne had higher self-esteem as compared to other grades of acne. Similarly, a cross-sectional study by Uslu et al¹⁵ found a direct correlation between acne severity and lower self-esteem using the RSES questionnaire. Although acne may be viewed as a minor cosmetic issue, it can have a negative impact on self-esteem and interpersonal relationships. Many of the studies had not used a validated structured questionnaire to assess self-esteem and there is a paucity of literature in relation to acne and self-esteem.^3,16,17

According to the WHOQOL-BREF, the psychological domain was affected more in cases than in controls, which was a statistically significant difference. One study observed that patients experience immediate psychological consequences of acne such as reduced self-esteem, poor self-image, self-consciousness, and embarrassment.³ These effects are exacerbated by taunting, stigmatization, and perceptions of scrutiny and being judged, causing patients to avoid interaction and social situations. Similarly, Pruthi and Babu¹⁸ observed that acne had an impact on the psychosocial aspects of adult females using the Dermatology Life Quality Index and CADI.

Financial resources, health and social care accessibility, and opportunities for acquiring new information and skills were the factors that were considered in the environment domain of the WHOQOL-BREF.⁸ We noted that the environment domain scores were significantly lower in cases than in controls. The cases could have had a detrimental effect on the latest opportunities in occupational functioning due to acne, and as most of the population was from a rural area, they were having less favorable circumstances in acquiring new information about the management of acne.

There was no statistically significant difference between cases and controls in the social and physical domains of the WHOQOL-BREF, which suggests that these fields do not influence QOL. Similarly, patients in Sarawak, Malaysia, were least affected in the domain of social functioning, which was likely attributed to the upbringing of this population encouraging stoicism.¹⁹

In the current study, QOL impairment showed a positive correlation with acne severity according to CADI scores; however, there was no significant difference between WHOQOL-BREF score and acne grading, which suggests that QOL impairment does not depend on severity of acne alone. Physical, psychological, social, and environment domains play an important role in impaired QOL. Hence, by using the WHOQOL-BREF we can evaluate the actual domain that is adversely affected by acne and can be treated with a holistic approach. This point must be stressed in the training of medical faculty, as the treatment of acne should not be based on acne severity alone but also on the degree of QOL impairment.¹⁹

These results indicate that more data are required and there is a need to consider other variables that could play a role. This study was a hospital-based, cross-sectional study with a small sample group that cannot be generalized, which are limitations. Longitudinal follow-up of the cases before and after treatment was not done. The questionnaires helped us to detect psychosocial aspects but were insufficient to diagnose psychiatric comorbidity.

The strengths of this study include the use of a specific scale for the assessment of self-esteem. The usage of comprehensive (WHOQOL-BREF) and specific (CADI) scales to evaluate QOL has mutual advantage.

Conclusion

Acne vulgaris is a disease that can adversely affect an individual’s QOL and self-esteem. This study suggested the importance of screening for psychosocial problems in those who present for management of acne. It is important for dermatologists to be cautious about psychological problems in acne patients and be aware of the importance of basic psychosomatic treatment in conjunction with medical treatment in the management of acne.

The incidence of skin cancer, along with its effects on patients and the economy, has continued to increase and therefore requires particular attention from dermatologists. UV light has been shown to be of etiopathologic importance in the development of various types of skin cancer.^1-3 Studies have shown that there is a direct correlation between the incidence of skin cancer and average annual amounts of UV radiation exposure.³ Accordingly, in 2009 the International Agency for Research on Cancer classified UV light as carcinogenic to humans.⁴ Therefore, the general public must be made aware of the danger of exposure to UV radiation.

In Australia, government initiatives to educate the population on causes of skin cancer development and its relationship to UV radiation have already caused the public to change their way of thinking and to deal with sunlight in a conscious and responsible manner.⁵ A large proportion of the Australian population with light skin is at a particularly high risk for developing skin cancer due to intense exposure to UV radiation. Numerous campaigns in Germany and other countries have attempted to sensitize the public to this issue by emphasizing a reduction in UV exposure (primary prevention) or highlighting the importance of early diagnosis (secondary prevention).^6,7

For a good prognosis, it is crucial that skin cancer, particularly melanoma, is discovered at an early or precancerous stage.⁸ For this reason, self-examination of the skin and skin cancer screening are important factors that can contribute to ensuring early and curative treatment.^9-11 Since July 1, 2008, skin cancer screenings have been included in the preventative health care program by statutory health insurance providers in Germany. As part of this program, the cost of screening once every 2 years for individuals 35 years and older is covered by statutory health insurance.¹² Several studies have shown a decline in the melanoma mortality rate since the introduction of skin cancer screening programs in Germany.^11,13,14

Employee skin cancer screening programs are an important method of examining high numbers of individuals quickly and effectively. These programs have been carried out in Germany and other countries.^15,16 Studies have shown that skin cancer screening carried out selectively on defined groups can be an effective form of secondary prevention, particularly for those who work outdoors.¹⁷

An employee skin cancer screening program was carried out as part of this study. The findings are interpreted and discussed in relation to other employee screening programs that have been reported as well as those introduced by statutory health insurance providers in Germany. The aim of this study was to determine the importance and effectiveness of employee skin cancer screening programs and the role they play in secondary prevention of skin cancer.

Methods

Study Population

Employees of a technical company in Bavaria, Germany, were offered a skin cancer screening program by the employer’s occupational health service and health insurance provider in collaboration with the Department of Dermatology at the University Hospital Erlangen (Erlangen, Germany). Skin examinations were performed exclusively by 5 trained dermatologists. Only direct employees of the company at 3 of its locations in the Erlangen area were eligible to participate. The total number of employees varied by location (1072–5126 employees). The majority of employees had a university education or had completed technical training. Family members and other individuals who were not members of the company were excluded. There were no further inclusion or exclusion criteria. Over a period of 13 days, 783 of 7823 total employees (10.0%) were examined and included in the study. The study was approved by the Responsible Ethics Commission of the Faculty of Medicine at Friedrich-Alexander-University Erlangen-Nürnberg, Germany.

Study Design

Employees signed a consent form for participation in the study and completed a standardized questionnaire. The questionnaire was based on surveys used in a prior study¹⁸ and collected information on current and prior skin lesions, prior dermatological screening, personal and family history of skin tumors, frequency of UV exposure, and type of UV protection used. For the question on measures taken for protection from UV radiation, possible answers included with sunscreen cream, with suitable sun-protective clothing, and by staying in the shade, or no measures were taken. In contrast to the other questions, multiple answers were accepted for this question. Answering no automatically excluded other possible answers. Participants also were asked to assess their own Fitzpatrick skin type¹⁹; the questionnaire included explanations of each skin type (I–IV).

The participants were then called in for examination by the dermatologist at 15-minute intervals. All clothing was removed and the skin was examined. Dermatoscopes were used for closer examination of suspicious skin lesions. The clinical results of the examinations were recorded on a standardized form.

An estimation of the number of melanocytic nevi—≤20, 21–49, or ≥50—was recorded for each patient. Suspicious skin lesions were assigned to one of the following categories: nevus requiring future checkup (Nc), nevus requiring excision (Ne), suspected malignant melanoma (MM), suspected squamous cell carcinoma, suspected basal cell carcinoma (BCC), suspected other skin tumor, and precancerous lesion. Fitzpatrick skin type also was assessed for all participants and recorded by the dermatologist carrying out the examination. Each participant was assigned to a risk group—low, moderate, or high risk—based on their individual risk for developing a skin tumor. Factors that were considered when determining participants’ risk for developing skin cancer included Fitzpatrick skin type, number of melanocytic nevi, personal and family history, leisure activities, UV protection used, and current clinical diagnosis of skin lesions.

After the skin examination, participants were informed of recommended treatment but were not given any additional dermatologic advice. Participants could arrange an appointment at the Department of Dermatology, University Hospital Erlangen, for the excision and histological analysis of the skin lesions. All recorded data were collected in a computerized spreadsheet program. When evaluating the questionnaires, questions that were not answered or were answered incorrectly (participant chose more than 1 answer) were ignored.

Statistical Analysis

Statistical analysis was carried out using SPSS software version 16.0. The majority of the data were nominal or ordinal. Metric data were checked for normal distribution using the Shapiro-Wilk test before carrying out parametric tests. Statistical tests were carried out using the χ² test and the t test for independent samples. Non-nominal distributed data were checked using the Mann-Whitney U test. P<.05 was considered statistically significant in the exploratory data analysis.

Results

Of 783 employees included in the study, 288 (36.8%) were female and 495 (63.2%) were male (Table 1). In comparison with the total workforce, a significantly higher proportion of women than men took part in the cross-sectional study (P<.01). The average age (SD) was 42.3 (9.5) years (range, 18–64 years). Female participants (average age [SD], 39.8 [10.2] years) were significantly younger than male participants (average age [SD], 43.8 [8.8] years; P<.01). Forty-one percent of participants had a prior skin cancer screening. One percent of participants had a personal history of skin cancer, with 1 participant reporting a history of MM; 6.5% had a family history of skin cancer, of which 39.2% had a family history of MM.

The results of the clinical examinations showed that 43.8% of participants had 20 or fewer melanocytic nevi, 43.4% had 21 to 49 melanocytic nevi, and 12.8% had 50 or more melanocytic nevi. Significantly more women than men had 20 or fewer melanocytic nevi (P<.05).

Approximately 92% of participants assessed themselves as having Fitzpatrick skin types II (35.2%) or III (56.7%), while only approximately 3.6% and 4.5% assessed themselves as having skin types I and IV, respectively. The results of the Fitzpatrick skin type assessments made by dermatologists were similar: 96.9% of participants were assessed as having Fitzpatrick skin types II (43.0%) and III (53.8%); approximately 1.9% and 1.3% were assessed as having Fitzpatrick skin types I and IV, respectively. Results showed that 80.2% of all participants assessed their skin type in the same way as the dermatologist; 13.5% assessed their skin type as darker and 6.3% (49/783) assessed it as lighter. A quantitative analysis of Fitzpatrick skin type and sex showed that significantly more male participants than female participants assessed their Fitzpatrick skin type darker than their actual skin type (P<.01).

Overall, 47.6% of participants reported having had sunburn rarely in the past, while 36.9% and 14.0% had experienced sunburn once per year and several times per year, respectively. Approximately 1.4% of participants reported never having a sunburn. More of the male participants made use of comprehensive sun protection using all methods listed (34.5%; P<.05) or a combination of sunscreen and sun-protective clothing (14.9%; P<.01) than the female participants who relied more frequently on sunscreen alone (29.5%; P<.01) or a combination of sunscreen and staying in the shade (29.5%; P<.01)

In general it was clear that sunscreen, either alone or in combination with other sun-protection methods, was used most frequently (88.0%); 58.0% protected themselves by staying in the shade, while 48.0% used suitable sun-protective clothing. Only 3.6% of participants did not protect themselves using any of the suggested methods.

A total of 661 categorized skin lesions were found in 377 participants. Of these lesions, 491 were Nc and 121 were Ne. Twenty-four of the skin lesions were suspected precancerous lesions, 13 were suspected BCC, 2 were suspected MM, and 10 were suspected other skin tumor (Table 2). Overall, male participants who were diagnosed with at least 1 skin lesion (average age, 44.0 years) were significantly older than the women (average age, 39.3 years)(P<.01). Similar findings were observed in participants with at least 1 Nc (men, 43.3 years; women, 38.7 years; P<.01) and at least 1 Ne (men, 44.2 years; women, 38.0 years; P<.05). With regard to the individual risk for developing skin cancer, 32.6% of participants were considered to be at low risk, 64.9% were at moderate risk, and 2.6% were at high risk.

Approximately 61.5% of 377 participants who were diagnosed with at least 1 categorized skin lesion were advised to have a specific skin lesion checked by a dermatologist or to have a full examination for skin cancer once every 12 months. Furthermore, 22.5% were advised to follow-up biannually and 11.7% were advised to follow-up once every 2 years. Of the remaining participants who were advised to have follow-ups, 0.3% were advised to have a skin examination once every 3 months after having had MM, and 4.0% were advised to have follow-up once every 18 months. Overall, follow-up was recommended within 1 year in 84.4% of cases and within 1 to 2 years in 15.6% (Table 3).

Subsequent histological analysis of the excised tissue resulted in a diagnosis of only 21 clinically significant skin conditions. One case of Bowen disease and 1 case of BCC was confirmed. Histological analysis identified the remaining 19 excised skin lesions, which included the 2 suspected MMs, as dysplastic nevi.

Comment

The aim of this cross-sectional study was to examine the importance and effectiveness of employee skin cancer screening programs. In comparison with the total workforce, significantly more women took part than men. Female participants were significantly younger than male participants, which mirrors the findings of prior studies showing that screening programs reach women more frequently than men and that women who participate in screenings are also younger on average in comparison to men.^7-13 Men and older individuals usually are underrepresented.^7,13 The average age of participants in our study was 42.3 years, which is lower than in the SCREEN (Skin Cancer Research to Provide Evidence for Effectiveness of Screening in Northern Germany) study (average age, 49.7 years).¹³ The average age in our study also is likely to be lower than patients who undergo skin cancer screenings offered by statutory health insurance providers in Germany, which has a minimum age restriction of 35 years; however, it is comparable to the average age of participants in other employee screening programs and therefore represents the average age of individuals employed in Germany.^15,16

The employee skin cancer screening program in this study generated a high level of interest, indicated by the fact that all available appointments had been booked just 36 hours after the screening was announced. Furthermore, there was a waiting list of approximately 300 employees who were not able to undergo a skin examination. For logistical reasons, the number of participants was limited to 10% of the workforce. The high level of interest is an indication of increased awareness of the importance of recognizing skin tumors early and the associated need for information as well as the need to undergo screening for skin cancer as a precaution. This observation also can be made with regard to the skin cancer screening introduced by statutory health insurance providers in Germany. Studies published by Augustin et al²⁰ and Kornek et al^21,22 confirm that skin cancer screenings have gained wide acceptance in Germany because they were introduced by statutory health insurance providers in 2008. The number of skin cancer screenings carried out by dermatologists in Germany also is increasing.^20-22 Although approximately 19% of those eligible to participate took part in the SCREEN pilot project,¹³ approximately 31% of individuals who were eligible to participate took part in skin cancer screenings offered by statutory health insurance providers in Germany in 2012, and the percentage is rising.²³ Two important factors affecting the high level of interest in the employee screening program used in our study were undoubtedly the advantages of the examination taking place during working hours and being held on the occupational health services’ premises in the workplace, which helped participants avoid the cost of travel and wait times associated with visiting a medical practice.

Of 783 participants included in this study, 377 displayed at least 1 categorized skin lesion; the majority were suspicious melanocytic nevi. This high incidence rate suggested that regular skin cancer screenings are useful, as it has been shown that there is a correlation between higher numbers of melanocytic nevi and increased risk for developing melanoma.²⁴

In a study by Winkler et al,²⁵ a skin cancer screening of 1658 bank and insurance employees found that 33.8% of those examined displayed at least 1 atypical melanocytic nevus and 27.2% displayed more than 50 melanocytic nevi (compared to 12.8% with ≥50 melanocytic nevi in the current study). The risk for developing skin cancer was classified as intermediate or high in 54.5% (compared to 67.5% at moderate or high risk in the current study).²⁵ Therefore, the rate of suspicious skin lesions was lower in the population of the study by Winkler et al²⁵ in comparison to the population of the current study. As the overall number of melanocytic nevi and the individual risk for skin cancer, however, was underestimated by the majority of the bank and insurance employees,²⁵ employee skin cancer screening programs can be used as a potentially effective tool to make employees aware of the issue and sensitizing them to it. Employee screening in addition to a final diagnosis can contribute to ensuring suitable treatment is started. For example, in the large-scale employee screening published by Schaefer et al¹⁵ and Augustin et al,¹⁶ 48,665 and 90,880 employees, respectively, were screened for inflammatory and noninflammatory skin diseases, and 19% and 27% of participants, respectively, were diagnosed with skin lesions that required treatment.

Participants in the current study were given no further treatment or advice. Recommendations were made that participants monitor suspicious skin lesions or have them removed. With regard to future screening, 84.4% of participants with at least 1 categorized skin lesion were advised to have a regular follow-up within 1 year, while 15.6% were advised to follow-up within 1 to 2 years. Therefore, a period of 2 years before the next checkup, the period between screenings offered by statutory health insurance providers in Germany,¹² was considered too long for the majority of participants, according to the dermatologists involved with our study.

Conclusion

The high rate of suspicious skin lesions diagnosed demonstrated the effectiveness of skin cancer screenings organized in the workplace, which should be recommended for all employees, not only those who are at high risk for developing skin cancer due to the nature of their work, such as those who work outdoors. It should be noted that the study group examined in the current study was a homogeneous group of employees of a technical company only and is therefore relatively selective. Nevertheless, despite the comparatively selective and young participant group, these examinations provide evidence of the importance of skin cancer screening programs for a wider population.

Acknowledgments

The authors thank Heidi Seybold, MD; Petra Wörl, MD; Sybille Thoma-Uszynski, MD; and Jens Bussmann, MD (all from Erlangen, Germany), for their support and active assistance in the practical implementation of this study.

The incidence of skin cancer, along with its effects on patients and the economy, has continued to increase and therefore requires particular attention from dermatologists. UV light has been shown to be of etiopathologic importance in the development of various types of skin cancer.^1-3 Studies have shown that there is a direct correlation between the incidence of skin cancer and average annual amounts of UV radiation exposure.³ Accordingly, in 2009 the International Agency for Research on Cancer classified UV light as carcinogenic to humans.⁴ Therefore, the general public must be made aware of the danger of exposure to UV radiation.

In Australia, government initiatives to educate the population on causes of skin cancer development and its relationship to UV radiation have already caused the public to change their way of thinking and to deal with sunlight in a conscious and responsible manner.⁵ A large proportion of the Australian population with light skin is at a particularly high risk for developing skin cancer due to intense exposure to UV radiation. Numerous campaigns in Germany and other countries have attempted to sensitize the public to this issue by emphasizing a reduction in UV exposure (primary prevention) or highlighting the importance of early diagnosis (secondary prevention).^6,7

For a good prognosis, it is crucial that skin cancer, particularly melanoma, is discovered at an early or precancerous stage.⁸ For this reason, self-examination of the skin and skin cancer screening are important factors that can contribute to ensuring early and curative treatment.^9-11 Since July 1, 2008, skin cancer screenings have been included in the preventative health care program by statutory health insurance providers in Germany. As part of this program, the cost of screening once every 2 years for individuals 35 years and older is covered by statutory health insurance.¹² Several studies have shown a decline in the melanoma mortality rate since the introduction of skin cancer screening programs in Germany.^11,13,14

Employee skin cancer screening programs are an important method of examining high numbers of individuals quickly and effectively. These programs have been carried out in Germany and other countries.^15,16 Studies have shown that skin cancer screening carried out selectively on defined groups can be an effective form of secondary prevention, particularly for those who work outdoors.¹⁷

An employee skin cancer screening program was carried out as part of this study. The findings are interpreted and discussed in relation to other employee screening programs that have been reported as well as those introduced by statutory health insurance providers in Germany. The aim of this study was to determine the importance and effectiveness of employee skin cancer screening programs and the role they play in secondary prevention of skin cancer.

Methods

Study Population

Employees of a technical company in Bavaria, Germany, were offered a skin cancer screening program by the employer’s occupational health service and health insurance provider in collaboration with the Department of Dermatology at the University Hospital Erlangen (Erlangen, Germany). Skin examinations were performed exclusively by 5 trained dermatologists. Only direct employees of the company at 3 of its locations in the Erlangen area were eligible to participate. The total number of employees varied by location (1072–5126 employees). The majority of employees had a university education or had completed technical training. Family members and other individuals who were not members of the company were excluded. There were no further inclusion or exclusion criteria. Over a period of 13 days, 783 of 7823 total employees (10.0%) were examined and included in the study. The study was approved by the Responsible Ethics Commission of the Faculty of Medicine at Friedrich-Alexander-University Erlangen-Nürnberg, Germany.

Study Design

Employees signed a consent form for participation in the study and completed a standardized questionnaire. The questionnaire was based on surveys used in a prior study¹⁸ and collected information on current and prior skin lesions, prior dermatological screening, personal and family history of skin tumors, frequency of UV exposure, and type of UV protection used. For the question on measures taken for protection from UV radiation, possible answers included with sunscreen cream, with suitable sun-protective clothing, and by staying in the shade, or no measures were taken. In contrast to the other questions, multiple answers were accepted for this question. Answering no automatically excluded other possible answers. Participants also were asked to assess their own Fitzpatrick skin type¹⁹; the questionnaire included explanations of each skin type (I–IV).

The participants were then called in for examination by the dermatologist at 15-minute intervals. All clothing was removed and the skin was examined. Dermatoscopes were used for closer examination of suspicious skin lesions. The clinical results of the examinations were recorded on a standardized form.

An estimation of the number of melanocytic nevi—≤20, 21–49, or ≥50—was recorded for each patient. Suspicious skin lesions were assigned to one of the following categories: nevus requiring future checkup (Nc), nevus requiring excision (Ne), suspected malignant melanoma (MM), suspected squamous cell carcinoma, suspected basal cell carcinoma (BCC), suspected other skin tumor, and precancerous lesion. Fitzpatrick skin type also was assessed for all participants and recorded by the dermatologist carrying out the examination. Each participant was assigned to a risk group—low, moderate, or high risk—based on their individual risk for developing a skin tumor. Factors that were considered when determining participants’ risk for developing skin cancer included Fitzpatrick skin type, number of melanocytic nevi, personal and family history, leisure activities, UV protection used, and current clinical diagnosis of skin lesions.

After the skin examination, participants were informed of recommended treatment but were not given any additional dermatologic advice. Participants could arrange an appointment at the Department of Dermatology, University Hospital Erlangen, for the excision and histological analysis of the skin lesions. All recorded data were collected in a computerized spreadsheet program. When evaluating the questionnaires, questions that were not answered or were answered incorrectly (participant chose more than 1 answer) were ignored.

Statistical Analysis

Statistical analysis was carried out using SPSS software version 16.0. The majority of the data were nominal or ordinal. Metric data were checked for normal distribution using the Shapiro-Wilk test before carrying out parametric tests. Statistical tests were carried out using the χ² test and the t test for independent samples. Non-nominal distributed data were checked using the Mann-Whitney U test. P<.05 was considered statistically significant in the exploratory data analysis.

Results

Of 783 employees included in the study, 288 (36.8%) were female and 495 (63.2%) were male (Table 1). In comparison with the total workforce, a significantly higher proportion of women than men took part in the cross-sectional study (P<.01). The average age (SD) was 42.3 (9.5) years (range, 18–64 years). Female participants (average age [SD], 39.8 [10.2] years) were significantly younger than male participants (average age [SD], 43.8 [8.8] years; P<.01). Forty-one percent of participants had a prior skin cancer screening. One percent of participants had a personal history of skin cancer, with 1 participant reporting a history of MM; 6.5% had a family history of skin cancer, of which 39.2% had a family history of MM.

The results of the clinical examinations showed that 43.8% of participants had 20 or fewer melanocytic nevi, 43.4% had 21 to 49 melanocytic nevi, and 12.8% had 50 or more melanocytic nevi. Significantly more women than men had 20 or fewer melanocytic nevi (P<.05).

Approximately 92% of participants assessed themselves as having Fitzpatrick skin types II (35.2%) or III (56.7%), while only approximately 3.6% and 4.5% assessed themselves as having skin types I and IV, respectively. The results of the Fitzpatrick skin type assessments made by dermatologists were similar: 96.9% of participants were assessed as having Fitzpatrick skin types II (43.0%) and III (53.8%); approximately 1.9% and 1.3% were assessed as having Fitzpatrick skin types I and IV, respectively. Results showed that 80.2% of all participants assessed their skin type in the same way as the dermatologist; 13.5% assessed their skin type as darker and 6.3% (49/783) assessed it as lighter. A quantitative analysis of Fitzpatrick skin type and sex showed that significantly more male participants than female participants assessed their Fitzpatrick skin type darker than their actual skin type (P<.01).

Overall, 47.6% of participants reported having had sunburn rarely in the past, while 36.9% and 14.0% had experienced sunburn once per year and several times per year, respectively. Approximately 1.4% of participants reported never having a sunburn. More of the male participants made use of comprehensive sun protection using all methods listed (34.5%; P<.05) or a combination of sunscreen and sun-protective clothing (14.9%; P<.01) than the female participants who relied more frequently on sunscreen alone (29.5%; P<.01) or a combination of sunscreen and staying in the shade (29.5%; P<.01)

In general it was clear that sunscreen, either alone or in combination with other sun-protection methods, was used most frequently (88.0%); 58.0% protected themselves by staying in the shade, while 48.0% used suitable sun-protective clothing. Only 3.6% of participants did not protect themselves using any of the suggested methods.

A total of 661 categorized skin lesions were found in 377 participants. Of these lesions, 491 were Nc and 121 were Ne. Twenty-four of the skin lesions were suspected precancerous lesions, 13 were suspected BCC, 2 were suspected MM, and 10 were suspected other skin tumor (Table 2). Overall, male participants who were diagnosed with at least 1 skin lesion (average age, 44.0 years) were significantly older than the women (average age, 39.3 years)(P<.01). Similar findings were observed in participants with at least 1 Nc (men, 43.3 years; women, 38.7 years; P<.01) and at least 1 Ne (men, 44.2 years; women, 38.0 years; P<.05). With regard to the individual risk for developing skin cancer, 32.6% of participants were considered to be at low risk, 64.9% were at moderate risk, and 2.6% were at high risk.

Approximately 61.5% of 377 participants who were diagnosed with at least 1 categorized skin lesion were advised to have a specific skin lesion checked by a dermatologist or to have a full examination for skin cancer once every 12 months. Furthermore, 22.5% were advised to follow-up biannually and 11.7% were advised to follow-up once every 2 years. Of the remaining participants who were advised to have follow-ups, 0.3% were advised to have a skin examination once every 3 months after having had MM, and 4.0% were advised to have follow-up once every 18 months. Overall, follow-up was recommended within 1 year in 84.4% of cases and within 1 to 2 years in 15.6% (Table 3).

Subsequent histological analysis of the excised tissue resulted in a diagnosis of only 21 clinically significant skin conditions. One case of Bowen disease and 1 case of BCC was confirmed. Histological analysis identified the remaining 19 excised skin lesions, which included the 2 suspected MMs, as dysplastic nevi.

Comment

The aim of this cross-sectional study was to examine the importance and effectiveness of employee skin cancer screening programs. In comparison with the total workforce, significantly more women took part than men. Female participants were significantly younger than male participants, which mirrors the findings of prior studies showing that screening programs reach women more frequently than men and that women who participate in screenings are also younger on average in comparison to men.^7-13 Men and older individuals usually are underrepresented.^7,13 The average age of participants in our study was 42.3 years, which is lower than in the SCREEN (Skin Cancer Research to Provide Evidence for Effectiveness of Screening in Northern Germany) study (average age, 49.7 years).¹³ The average age in our study also is likely to be lower than patients who undergo skin cancer screenings offered by statutory health insurance providers in Germany, which has a minimum age restriction of 35 years; however, it is comparable to the average age of participants in other employee screening programs and therefore represents the average age of individuals employed in Germany.^15,16

The employee skin cancer screening program in this study generated a high level of interest, indicated by the fact that all available appointments had been booked just 36 hours after the screening was announced. Furthermore, there was a waiting list of approximately 300 employees who were not able to undergo a skin examination. For logistical reasons, the number of participants was limited to 10% of the workforce. The high level of interest is an indication of increased awareness of the importance of recognizing skin tumors early and the associated need for information as well as the need to undergo screening for skin cancer as a precaution. This observation also can be made with regard to the skin cancer screening introduced by statutory health insurance providers in Germany. Studies published by Augustin et al²⁰ and Kornek et al^21,22 confirm that skin cancer screenings have gained wide acceptance in Germany because they were introduced by statutory health insurance providers in 2008. The number of skin cancer screenings carried out by dermatologists in Germany also is increasing.^20-22 Although approximately 19% of those eligible to participate took part in the SCREEN pilot project,¹³ approximately 31% of individuals who were eligible to participate took part in skin cancer screenings offered by statutory health insurance providers in Germany in 2012, and the percentage is rising.²³ Two important factors affecting the high level of interest in the employee screening program used in our study were undoubtedly the advantages of the examination taking place during working hours and being held on the occupational health services’ premises in the workplace, which helped participants avoid the cost of travel and wait times associated with visiting a medical practice.

Of 783 participants included in this study, 377 displayed at least 1 categorized skin lesion; the majority were suspicious melanocytic nevi. This high incidence rate suggested that regular skin cancer screenings are useful, as it has been shown that there is a correlation between higher numbers of melanocytic nevi and increased risk for developing melanoma.²⁴

In a study by Winkler et al,²⁵ a skin cancer screening of 1658 bank and insurance employees found that 33.8% of those examined displayed at least 1 atypical melanocytic nevus and 27.2% displayed more than 50 melanocytic nevi (compared to 12.8% with ≥50 melanocytic nevi in the current study). The risk for developing skin cancer was classified as intermediate or high in 54.5% (compared to 67.5% at moderate or high risk in the current study).²⁵ Therefore, the rate of suspicious skin lesions was lower in the population of the study by Winkler et al²⁵ in comparison to the population of the current study. As the overall number of melanocytic nevi and the individual risk for skin cancer, however, was underestimated by the majority of the bank and insurance employees,²⁵ employee skin cancer screening programs can be used as a potentially effective tool to make employees aware of the issue and sensitizing them to it. Employee screening in addition to a final diagnosis can contribute to ensuring suitable treatment is started. For example, in the large-scale employee screening published by Schaefer et al¹⁵ and Augustin et al,¹⁶ 48,665 and 90,880 employees, respectively, were screened for inflammatory and noninflammatory skin diseases, and 19% and 27% of participants, respectively, were diagnosed with skin lesions that required treatment.

Participants in the current study were given no further treatment or advice. Recommendations were made that participants monitor suspicious skin lesions or have them removed. With regard to future screening, 84.4% of participants with at least 1 categorized skin lesion were advised to have a regular follow-up within 1 year, while 15.6% were advised to follow-up within 1 to 2 years. Therefore, a period of 2 years before the next checkup, the period between screenings offered by statutory health insurance providers in Germany,¹² was considered too long for the majority of participants, according to the dermatologists involved with our study.

Conclusion

The high rate of suspicious skin lesions diagnosed demonstrated the effectiveness of skin cancer screenings organized in the workplace, which should be recommended for all employees, not only those who are at high risk for developing skin cancer due to the nature of their work, such as those who work outdoors. It should be noted that the study group examined in the current study was a homogeneous group of employees of a technical company only and is therefore relatively selective. Nevertheless, despite the comparatively selective and young participant group, these examinations provide evidence of the importance of skin cancer screening programs for a wider population.

Acknowledgments

The authors thank Heidi Seybold, MD; Petra Wörl, MD; Sybille Thoma-Uszynski, MD; and Jens Bussmann, MD (all from Erlangen, Germany), for their support and active assistance in the practical implementation of this study.

The incidence of skin cancer, along with its effects on patients and the economy, has continued to increase and therefore requires particular attention from dermatologists. UV light has been shown to be of etiopathologic importance in the development of various types of skin cancer.^1-3 Studies have shown that there is a direct correlation between the incidence of skin cancer and average annual amounts of UV radiation exposure.³ Accordingly, in 2009 the International Agency for Research on Cancer classified UV light as carcinogenic to humans.⁴ Therefore, the general public must be made aware of the danger of exposure to UV radiation.

In Australia, government initiatives to educate the population on causes of skin cancer development and its relationship to UV radiation have already caused the public to change their way of thinking and to deal with sunlight in a conscious and responsible manner.⁵ A large proportion of the Australian population with light skin is at a particularly high risk for developing skin cancer due to intense exposure to UV radiation. Numerous campaigns in Germany and other countries have attempted to sensitize the public to this issue by emphasizing a reduction in UV exposure (primary prevention) or highlighting the importance of early diagnosis (secondary prevention).^6,7

For a good prognosis, it is crucial that skin cancer, particularly melanoma, is discovered at an early or precancerous stage.⁸ For this reason, self-examination of the skin and skin cancer screening are important factors that can contribute to ensuring early and curative treatment.^9-11 Since July 1, 2008, skin cancer screenings have been included in the preventative health care program by statutory health insurance providers in Germany. As part of this program, the cost of screening once every 2 years for individuals 35 years and older is covered by statutory health insurance.¹² Several studies have shown a decline in the melanoma mortality rate since the introduction of skin cancer screening programs in Germany.^11,13,14

Employee skin cancer screening programs are an important method of examining high numbers of individuals quickly and effectively. These programs have been carried out in Germany and other countries.^15,16 Studies have shown that skin cancer screening carried out selectively on defined groups can be an effective form of secondary prevention, particularly for those who work outdoors.¹⁷

An employee skin cancer screening program was carried out as part of this study. The findings are interpreted and discussed in relation to other employee screening programs that have been reported as well as those introduced by statutory health insurance providers in Germany. The aim of this study was to determine the importance and effectiveness of employee skin cancer screening programs and the role they play in secondary prevention of skin cancer.

Methods

Study Population

Employees of a technical company in Bavaria, Germany, were offered a skin cancer screening program by the employer’s occupational health service and health insurance provider in collaboration with the Department of Dermatology at the University Hospital Erlangen (Erlangen, Germany). Skin examinations were performed exclusively by 5 trained dermatologists. Only direct employees of the company at 3 of its locations in the Erlangen area were eligible to participate. The total number of employees varied by location (1072–5126 employees). The majority of employees had a university education or had completed technical training. Family members and other individuals who were not members of the company were excluded. There were no further inclusion or exclusion criteria. Over a period of 13 days, 783 of 7823 total employees (10.0%) were examined and included in the study. The study was approved by the Responsible Ethics Commission of the Faculty of Medicine at Friedrich-Alexander-University Erlangen-Nürnberg, Germany.

Study Design

Employees signed a consent form for participation in the study and completed a standardized questionnaire. The questionnaire was based on surveys used in a prior study¹⁸ and collected information on current and prior skin lesions, prior dermatological screening, personal and family history of skin tumors, frequency of UV exposure, and type of UV protection used. For the question on measures taken for protection from UV radiation, possible answers included with sunscreen cream, with suitable sun-protective clothing, and by staying in the shade, or no measures were taken. In contrast to the other questions, multiple answers were accepted for this question. Answering no automatically excluded other possible answers. Participants also were asked to assess their own Fitzpatrick skin type¹⁹; the questionnaire included explanations of each skin type (I–IV).

The participants were then called in for examination by the dermatologist at 15-minute intervals. All clothing was removed and the skin was examined. Dermatoscopes were used for closer examination of suspicious skin lesions. The clinical results of the examinations were recorded on a standardized form.

An estimation of the number of melanocytic nevi—≤20, 21–49, or ≥50—was recorded for each patient. Suspicious skin lesions were assigned to one of the following categories: nevus requiring future checkup (Nc), nevus requiring excision (Ne), suspected malignant melanoma (MM), suspected squamous cell carcinoma, suspected basal cell carcinoma (BCC), suspected other skin tumor, and precancerous lesion. Fitzpatrick skin type also was assessed for all participants and recorded by the dermatologist carrying out the examination. Each participant was assigned to a risk group—low, moderate, or high risk—based on their individual risk for developing a skin tumor. Factors that were considered when determining participants’ risk for developing skin cancer included Fitzpatrick skin type, number of melanocytic nevi, personal and family history, leisure activities, UV protection used, and current clinical diagnosis of skin lesions.

After the skin examination, participants were informed of recommended treatment but were not given any additional dermatologic advice. Participants could arrange an appointment at the Department of Dermatology, University Hospital Erlangen, for the excision and histological analysis of the skin lesions. All recorded data were collected in a computerized spreadsheet program. When evaluating the questionnaires, questions that were not answered or were answered incorrectly (participant chose more than 1 answer) were ignored.

Statistical Analysis

Statistical analysis was carried out using SPSS software version 16.0. The majority of the data were nominal or ordinal. Metric data were checked for normal distribution using the Shapiro-Wilk test before carrying out parametric tests. Statistical tests were carried out using the χ² test and the t test for independent samples. Non-nominal distributed data were checked using the Mann-Whitney U test. P<.05 was considered statistically significant in the exploratory data analysis.

Results

Of 783 employees included in the study, 288 (36.8%) were female and 495 (63.2%) were male (Table 1). In comparison with the total workforce, a significantly higher proportion of women than men took part in the cross-sectional study (P<.01). The average age (SD) was 42.3 (9.5) years (range, 18–64 years). Female participants (average age [SD], 39.8 [10.2] years) were significantly younger than male participants (average age [SD], 43.8 [8.8] years; P<.01). Forty-one percent of participants had a prior skin cancer screening. One percent of participants had a personal history of skin cancer, with 1 participant reporting a history of MM; 6.5% had a family history of skin cancer, of which 39.2% had a family history of MM.

The results of the clinical examinations showed that 43.8% of participants had 20 or fewer melanocytic nevi, 43.4% had 21 to 49 melanocytic nevi, and 12.8% had 50 or more melanocytic nevi. Significantly more women than men had 20 or fewer melanocytic nevi (P<.05).

Approximately 92% of participants assessed themselves as having Fitzpatrick skin types II (35.2%) or III (56.7%), while only approximately 3.6% and 4.5% assessed themselves as having skin types I and IV, respectively. The results of the Fitzpatrick skin type assessments made by dermatologists were similar: 96.9% of participants were assessed as having Fitzpatrick skin types II (43.0%) and III (53.8%); approximately 1.9% and 1.3% were assessed as having Fitzpatrick skin types I and IV, respectively. Results showed that 80.2% of all participants assessed their skin type in the same way as the dermatologist; 13.5% assessed their skin type as darker and 6.3% (49/783) assessed it as lighter. A quantitative analysis of Fitzpatrick skin type and sex showed that significantly more male participants than female participants assessed their Fitzpatrick skin type darker than their actual skin type (P<.01).

Overall, 47.6% of participants reported having had sunburn rarely in the past, while 36.9% and 14.0% had experienced sunburn once per year and several times per year, respectively. Approximately 1.4% of participants reported never having a sunburn. More of the male participants made use of comprehensive sun protection using all methods listed (34.5%; P<.05) or a combination of sunscreen and sun-protective clothing (14.9%; P<.01) than the female participants who relied more frequently on sunscreen alone (29.5%; P<.01) or a combination of sunscreen and staying in the shade (29.5%; P<.01)

In general it was clear that sunscreen, either alone or in combination with other sun-protection methods, was used most frequently (88.0%); 58.0% protected themselves by staying in the shade, while 48.0% used suitable sun-protective clothing. Only 3.6% of participants did not protect themselves using any of the suggested methods.

A total of 661 categorized skin lesions were found in 377 participants. Of these lesions, 491 were Nc and 121 were Ne. Twenty-four of the skin lesions were suspected precancerous lesions, 13 were suspected BCC, 2 were suspected MM, and 10 were suspected other skin tumor (Table 2). Overall, male participants who were diagnosed with at least 1 skin lesion (average age, 44.0 years) were significantly older than the women (average age, 39.3 years)(P<.01). Similar findings were observed in participants with at least 1 Nc (men, 43.3 years; women, 38.7 years; P<.01) and at least 1 Ne (men, 44.2 years; women, 38.0 years; P<.05). With regard to the individual risk for developing skin cancer, 32.6% of participants were considered to be at low risk, 64.9% were at moderate risk, and 2.6% were at high risk.

Approximately 61.5% of 377 participants who were diagnosed with at least 1 categorized skin lesion were advised to have a specific skin lesion checked by a dermatologist or to have a full examination for skin cancer once every 12 months. Furthermore, 22.5% were advised to follow-up biannually and 11.7% were advised to follow-up once every 2 years. Of the remaining participants who were advised to have follow-ups, 0.3% were advised to have a skin examination once every 3 months after having had MM, and 4.0% were advised to have follow-up once every 18 months. Overall, follow-up was recommended within 1 year in 84.4% of cases and within 1 to 2 years in 15.6% (Table 3).

Subsequent histological analysis of the excised tissue resulted in a diagnosis of only 21 clinically significant skin conditions. One case of Bowen disease and 1 case of BCC was confirmed. Histological analysis identified the remaining 19 excised skin lesions, which included the 2 suspected MMs, as dysplastic nevi.

Comment

The aim of this cross-sectional study was to examine the importance and effectiveness of employee skin cancer screening programs. In comparison with the total workforce, significantly more women took part than men. Female participants were significantly younger than male participants, which mirrors the findings of prior studies showing that screening programs reach women more frequently than men and that women who participate in screenings are also younger on average in comparison to men.^7-13 Men and older individuals usually are underrepresented.^7,13 The average age of participants in our study was 42.3 years, which is lower than in the SCREEN (Skin Cancer Research to Provide Evidence for Effectiveness of Screening in Northern Germany) study (average age, 49.7 years).¹³ The average age in our study also is likely to be lower than patients who undergo skin cancer screenings offered by statutory health insurance providers in Germany, which has a minimum age restriction of 35 years; however, it is comparable to the average age of participants in other employee screening programs and therefore represents the average age of individuals employed in Germany.^15,16

The employee skin cancer screening program in this study generated a high level of interest, indicated by the fact that all available appointments had been booked just 36 hours after the screening was announced. Furthermore, there was a waiting list of approximately 300 employees who were not able to undergo a skin examination. For logistical reasons, the number of participants was limited to 10% of the workforce. The high level of interest is an indication of increased awareness of the importance of recognizing skin tumors early and the associated need for information as well as the need to undergo screening for skin cancer as a precaution. This observation also can be made with regard to the skin cancer screening introduced by statutory health insurance providers in Germany. Studies published by Augustin et al²⁰ and Kornek et al^21,22 confirm that skin cancer screenings have gained wide acceptance in Germany because they were introduced by statutory health insurance providers in 2008. The number of skin cancer screenings carried out by dermatologists in Germany also is increasing.^20-22 Although approximately 19% of those eligible to participate took part in the SCREEN pilot project,¹³ approximately 31% of individuals who were eligible to participate took part in skin cancer screenings offered by statutory health insurance providers in Germany in 2012, and the percentage is rising.²³ Two important factors affecting the high level of interest in the employee screening program used in our study were undoubtedly the advantages of the examination taking place during working hours and being held on the occupational health services’ premises in the workplace, which helped participants avoid the cost of travel and wait times associated with visiting a medical practice.

Of 783 participants included in this study, 377 displayed at least 1 categorized skin lesion; the majority were suspicious melanocytic nevi. This high incidence rate suggested that regular skin cancer screenings are useful, as it has been shown that there is a correlation between higher numbers of melanocytic nevi and increased risk for developing melanoma.²⁴

In a study by Winkler et al,²⁵ a skin cancer screening of 1658 bank and insurance employees found that 33.8% of those examined displayed at least 1 atypical melanocytic nevus and 27.2% displayed more than 50 melanocytic nevi (compared to 12.8% with ≥50 melanocytic nevi in the current study). The risk for developing skin cancer was classified as intermediate or high in 54.5% (compared to 67.5% at moderate or high risk in the current study).²⁵ Therefore, the rate of suspicious skin lesions was lower in the population of the study by Winkler et al²⁵ in comparison to the population of the current study. As the overall number of melanocytic nevi and the individual risk for skin cancer, however, was underestimated by the majority of the bank and insurance employees,²⁵ employee skin cancer screening programs can be used as a potentially effective tool to make employees aware of the issue and sensitizing them to it. Employee screening in addition to a final diagnosis can contribute to ensuring suitable treatment is started. For example, in the large-scale employee screening published by Schaefer et al¹⁵ and Augustin et al,¹⁶ 48,665 and 90,880 employees, respectively, were screened for inflammatory and noninflammatory skin diseases, and 19% and 27% of participants, respectively, were diagnosed with skin lesions that required treatment.

Participants in the current study were given no further treatment or advice. Recommendations were made that participants monitor suspicious skin lesions or have them removed. With regard to future screening, 84.4% of participants with at least 1 categorized skin lesion were advised to have a regular follow-up within 1 year, while 15.6% were advised to follow-up within 1 to 2 years. Therefore, a period of 2 years before the next checkup, the period between screenings offered by statutory health insurance providers in Germany,¹² was considered too long for the majority of participants, according to the dermatologists involved with our study.

Conclusion

The high rate of suspicious skin lesions diagnosed demonstrated the effectiveness of skin cancer screenings organized in the workplace, which should be recommended for all employees, not only those who are at high risk for developing skin cancer due to the nature of their work, such as those who work outdoors. It should be noted that the study group examined in the current study was a homogeneous group of employees of a technical company only and is therefore relatively selective. Nevertheless, despite the comparatively selective and young participant group, these examinations provide evidence of the importance of skin cancer screening programs for a wider population.

Acknowledgments

The authors thank Heidi Seybold, MD; Petra Wörl, MD; Sybille Thoma-Uszynski, MD; and Jens Bussmann, MD (all from Erlangen, Germany), for their support and active assistance in the practical implementation of this study.

Although rare, periprosthetic fractures remain a significant complication after total knee arthroplasty (TKA), occurring in 0.3% to 2.5% of cases.^1-4 Hirsh and colleagues⁵ were among the first to suggest that anterior femoral notching during TKA was a potential risk factor for postoperative periprosthetic femoral fracture because notching may weaken the anterior femoral cortex. Anterior femoral notching, a cortex violation occurring during an anterior bone cut, occurs in up to 30% of cases.⁶ Using a theoretical biomechanical model, Culp and colleagues¹ found that increasing the depth of the notch defect into the cortex led to reduced torsional strength. In more recent, cadaveric biomechanical studies, notching of the anterior femoral cortex decreased torsional strength by up to 39%.^7,8 Contrary to these biomechanical studies, a retrospective study evaluating 1089 TKAs using 2 implant designs (Anatomic Graduated Component, Biomet and Legacy, Zimmer) demonstrated no significant effect of anterior femoral notching with respect to incidence of supracondylar femur fractures.⁶ That study, however, did not address whether implant design is associated with a differential risk for fracture in the presence of anterior notching.

Previous biomechanical studies have primarily investigated cruciate-retaining (CR) femoral components and properties with respect to anterior notching, even though the posterior-stabilized (PS) design is used more often in the United States.^1,7 According to a Mayo Clinic survey, TKAs with a PS design increased from <10% in 1990 to almost 75% by 1997.⁹ Today, there is little to no consensus about which implant is better, and use of one or the other depends largely on the surgeon and varies widely between countries and regions.¹⁰ PS designs require more bone resection and demonstrate prosthesis-controlled rollback during flexion, whereas CR designs preserve more bone and achieve posterior stabilization via the posterior cruciate ligament.¹¹ Despite these differences in design and mechanics, a 2013 Cochrane review of TKA design found no clinically significant differences between CR and PS with respect to pain, range of motion, or clinical and radiologic outcomes.¹⁰ The reviewers did not specifically address periprosthetic fractures associated with either femoral notching or TKA design, as they could not quantitatively analyze postoperative complications because of the diversity of reports. Given the limited number of reported cases, a review of radiographic findings pertaining to the characteristics of supracondylar fractures in anterior femoral notching was unsuccessful.¹² As the previous biomechanical studies of anterior notching used primarily CR models or no prostheses at all, a study of biomechanical differences between CR and PS designs in the presence of anterior notching is warranted.^1,7,8 Therefore, we conducted a study to assess the effect of anterior femoral notching on torsional strength and load to failure in CR and PS femoral components.

Materials and Methods

Twelve fourth-generation composite adult left femur synthetic sawbones (Sawbones; Pacific Research Laboratories) were selected for their consistent biomechanical properties, vs those of cadaveric specimens; in addition, low intersample variability made them preferable to cadaveric bones given the small sample used in this study.^13,14 All bones were from the same lot. All were visually inspected for defects and found to be acceptable. In each sample, an anterior cortical defect was created by making an anterior cut with an undersized (size 4) posterior referencing guide. In addition, the distance from the proximal end of the notch to the implant fell within 15 mm, as that is the maximum distance from the implant a notch can be placed using a standard femoral cutting jig.¹⁵ Six femora were instrumented with CR implants and 6 with PS implants (DePuy Synthes). Implants were placed using standardized cuts. Before testing, each implant was inspected for proper fit and found to be securely fastened to the femur. In addition, precision calipers were used to measure notch depth and distance from notch to implant before loading. A custom polymethylmethacrylate torsion jig was used to fix each instrumented femur proximally and distally on the femoral implant (Figure 1). Care was taken to ensure the distal jig engaged only the implant, thus isolating the notch as a stress riser. Each femur was loaded in external rotation through the proximal femoral jig along the anatomical axis. Use of external rotation was based on study findings implicating external rotation of the tibia as the most likely mechanism for generating a fracture in the event of a fall.¹² Furthermore, distal femur fractures are predominantly spiral as opposed to butterfly or bending—an indication that torsion is the most likely mechanism of failure.¹⁶ With no axial rotation possible within the prosthesis, increased torsional stress is undoubtedly generated within adjacent bone. Each specimen underwent torsional stiffness testing and then load to failure. Torsional stiffness was measured by slowly loading each femur in external rotation, from 1 to 18 Nm for 3 cycles at a displacement rate of 0.5° per second. Each specimen then underwent torsional load-to-failure testing on an Instron 5800R machine at a rate of 0.5° per second. Failure was defined as the moment of fracture and subsequent decrease in torsional load—determined graphically by the peak torsional load followed immediately by a sharp decrease in load. Stiffness was determined as the slope of torque to the displacement curve for each cycle, and torque to failure was the highest recorded torque before fracture. Fracture pattern was noted after failure. A sample size of 6 specimens per group provided 80% power to detect a between-group difference of 1 Nm per degree in stiffness, using an estimated SD of 0.7 Nm per degree. In our statistical analysis, continuous variables are reported as means and SDs. Data from our torsional stiffness and load-to-failure testing were analyzed with unpaired 2-sample t tests, and P < .05 was considered statistically significant.

Results

We did not detect a statistical difference in notch depth, notch-to-implant distance, or femoral length between the CR and PS groups. Mean (SD) notch depth was 6.0 (1.3) mm for CR and 4.9 (1.0) mm for PS (P = .13); mean (SD) distance from the proximal end of the notch to the implant was 13.8 (1.7) mm for CR and 11.1 (3.2) mm for PS (P = .08); and mean (SD) femoral length was 46.2 (0.1) cm for CR and 46.2 (0.1) cm for PS (P = .60).

Mean (SD) torsional stiffness for the first 3 precycles was 6.2 (1.2), 8.7 (1.5), and 8.8 (1.4) Nm per degree for the CR group and 6.0 (0.7), 8.4 (1.4), and 8.6 (1.4) Nm per degree for the PS group; the differences were not statistically significant (Figure 2A). In addition, there were no statistically significant differences in mean (SD) stiffness at failure between CR, 6.5 (0.7) Nm per degree, and PS, 7.1 (0.9) Nm per degree (P = .24; Figure 2B) or in mean (SD) final torque at failure between CR, 62.4 (9.4) Nm, and PS, 62.7 (12.2) Nm (P = .95; Figure 2C).

All fractures in both groups were oblique fractures originating at the proximal angle of the notch and extended proximally. None extended distally into the box. Fracture locations and patterns were identical in the CR and PS groups of femurs (Figure 3).

Discussion

Periprosthetic fractures after TKA remain rare. However, these fractures can significantly increase morbidity and complications. Anterior femoral notching occurs inadvertently in 30% to 40% of TKAs.^6,17 The impact of femoral notching on supracondylar femur fracture is inconsistent between biomechanical and retrospective clinical studies. Retrospective studies failed to find a significant correlation between anterior femoral notching and supracondylar femur fractures.^6,17 However, findings of biomechanical studies have suggested that a notch 3 mm deep will reduce the torsional strength of the femur by 29%.⁷ Another study, using 3-dimensional finite element analysis, showed a significant increase in local stress with a notch deeper than 3 mm.¹⁵

To our knowledge, no clinical studies, including the aforementioned Cochrane review,¹⁰ have specifically evaluated the difference in risk for periprosthetic fracture between different TKA models in the presence of notching.¹¹ The biomechanical differences between implant designs could be a confounding factor in the results of past studies. More bone resection is required in PS designs than in CR designs. The position of the PS intercondylar cutout, much lower than the top of the patella flange, should not increase susceptibility to fractures more than in CR designs, but this hypothesis, though accepted, has not been validated biomechanically or addressed specifically in prospective or retrospective clinical analysis. In the present study, we used a biomechanical model to replicate an external rotation failure mechanism and quantify the differences in torsional strength and load to failure between CR TKA and PS TKA models in the presence of anterior femoral notching. Our results showed no significant differences in torsional stiffness, stiffness at failure, or torque at failure between the CR and PS design groups in the presence of anterior femoral notching.

In this study, all femoral fractures were oblique, and they all originated at the site of the cortical defect, not the notch—a situation markedly different from having bending forces applied to the femur. Previous biomechanical data indicated that bending forces applied to a notched femur cause fractures originating at the notch, whereas torsional forces applied to a notched femur cause fractures originating at the anterior aspect of the bone–component interface.⁷ The difference is attributable to study design. Our femurs were held fixed at their proximal end, which may have exacerbated any bending forces applied during external rotation, but we thought constraining the proximal femur would better replicate a fall involving external rotation.

More important for our study, an oblique fracture pattern was noted for both design groups (CR and PS), indicating the fracture pattern was unrelated to the area from which bone was resected for the PS design. All femur fractures in both design groups occurred proximal to a well-fixed prosthesis, indicating they should be classified as Vancouver C fractures. This is significant because intercondylar fossa resection (PS group) did not convert the fractures into Vancouver B2 fractures, which involve prosthesis loosening caused by pericomponent fracture.¹⁸ This simple observation validated our hypothesis that there would be no biomechanical differences between CR and PS designs with respect to the effects of anterior femoral notching. This lack of a significant difference may be attributed to the PS intercondylar cutout being much lower than the top of the anterior flange shielding the resected bone deep to the anterior flange.⁷ In addition, given the rarity of supracondylar fractures and the lack of sufficient relevant clinical data, it is difficult to speculate on the fracture patterns observed in clinical cases versus biomechanical studies.¹²

The use of synthetic bone models instead of cadaveric specimens could be seen as a limitation. Although synthetic bones may not reproduce the mechanism of failure in living and cadaveric femurs, the mechanical properties of synthetic bones have previously been found to fall within the range of those of cadaveric bones under axial loading, bending, and torsion testing.^13,14 As a uniform testing material, synthetic bones allow removal of the confounding variations in bone size and quality that plague biomechanical studies in cadaveric bones.^13,14 Interfemoral variability was 20 to 200 times higher in cadaveric femurs than in synthetic bones, which makes synthetic femurs preferable to cadaveric femurs, especially in studies with a small sample size.^13,14 In addition, a uniform specimen provides consistent, reproducible osteotomies, which were crucial for consistent mechanical evaluation of each configuration in this study.

The long-term clinical significance of anterior femoral notching in periprosthetic fractures is equivocal, possibly because most studies predominantly use CR implants.⁶ This may not be an issue if it is shown that CR and PS implants have the same mechanical properties. Despite the differences between clinical studies and our biomechanical study, reevaluation of clinical data is not warranted given the biomechanical data we present here. Results of biomechanical studies like ours still suggest an increased immediate postoperative risk for supracondylar fracture after anterior cortical notching of the femur.^5,7 Ultimately, this study found that, compared with a CR design, a PS design did not alter the torsional biomechanical properties or fracture pattern of an anteriorly notched femur.

Although rare, periprosthetic fractures remain a significant complication after total knee arthroplasty (TKA), occurring in 0.3% to 2.5% of cases.^1-4 Hirsh and colleagues⁵ were among the first to suggest that anterior femoral notching during TKA was a potential risk factor for postoperative periprosthetic femoral fracture because notching may weaken the anterior femoral cortex. Anterior femoral notching, a cortex violation occurring during an anterior bone cut, occurs in up to 30% of cases.⁶ Using a theoretical biomechanical model, Culp and colleagues¹ found that increasing the depth of the notch defect into the cortex led to reduced torsional strength. In more recent, cadaveric biomechanical studies, notching of the anterior femoral cortex decreased torsional strength by up to 39%.^7,8 Contrary to these biomechanical studies, a retrospective study evaluating 1089 TKAs using 2 implant designs (Anatomic Graduated Component, Biomet and Legacy, Zimmer) demonstrated no significant effect of anterior femoral notching with respect to incidence of supracondylar femur fractures.⁶ That study, however, did not address whether implant design is associated with a differential risk for fracture in the presence of anterior notching.

Previous biomechanical studies have primarily investigated cruciate-retaining (CR) femoral components and properties with respect to anterior notching, even though the posterior-stabilized (PS) design is used more often in the United States.^1,7 According to a Mayo Clinic survey, TKAs with a PS design increased from <10% in 1990 to almost 75% by 1997.⁹ Today, there is little to no consensus about which implant is better, and use of one or the other depends largely on the surgeon and varies widely between countries and regions.¹⁰ PS designs require more bone resection and demonstrate prosthesis-controlled rollback during flexion, whereas CR designs preserve more bone and achieve posterior stabilization via the posterior cruciate ligament.¹¹ Despite these differences in design and mechanics, a 2013 Cochrane review of TKA design found no clinically significant differences between CR and PS with respect to pain, range of motion, or clinical and radiologic outcomes.¹⁰ The reviewers did not specifically address periprosthetic fractures associated with either femoral notching or TKA design, as they could not quantitatively analyze postoperative complications because of the diversity of reports. Given the limited number of reported cases, a review of radiographic findings pertaining to the characteristics of supracondylar fractures in anterior femoral notching was unsuccessful.¹² As the previous biomechanical studies of anterior notching used primarily CR models or no prostheses at all, a study of biomechanical differences between CR and PS designs in the presence of anterior notching is warranted.^1,7,8 Therefore, we conducted a study to assess the effect of anterior femoral notching on torsional strength and load to failure in CR and PS femoral components.

Materials and Methods

Twelve fourth-generation composite adult left femur synthetic sawbones (Sawbones; Pacific Research Laboratories) were selected for their consistent biomechanical properties, vs those of cadaveric specimens; in addition, low intersample variability made them preferable to cadaveric bones given the small sample used in this study.^13,14 All bones were from the same lot. All were visually inspected for defects and found to be acceptable. In each sample, an anterior cortical defect was created by making an anterior cut with an undersized (size 4) posterior referencing guide. In addition, the distance from the proximal end of the notch to the implant fell within 15 mm, as that is the maximum distance from the implant a notch can be placed using a standard femoral cutting jig.¹⁵ Six femora were instrumented with CR implants and 6 with PS implants (DePuy Synthes). Implants were placed using standardized cuts. Before testing, each implant was inspected for proper fit and found to be securely fastened to the femur. In addition, precision calipers were used to measure notch depth and distance from notch to implant before loading. A custom polymethylmethacrylate torsion jig was used to fix each instrumented femur proximally and distally on the femoral implant (Figure 1). Care was taken to ensure the distal jig engaged only the implant, thus isolating the notch as a stress riser. Each femur was loaded in external rotation through the proximal femoral jig along the anatomical axis. Use of external rotation was based on study findings implicating external rotation of the tibia as the most likely mechanism for generating a fracture in the event of a fall.¹² Furthermore, distal femur fractures are predominantly spiral as opposed to butterfly or bending—an indication that torsion is the most likely mechanism of failure.¹⁶ With no axial rotation possible within the prosthesis, increased torsional stress is undoubtedly generated within adjacent bone. Each specimen underwent torsional stiffness testing and then load to failure. Torsional stiffness was measured by slowly loading each femur in external rotation, from 1 to 18 Nm for 3 cycles at a displacement rate of 0.5° per second. Each specimen then underwent torsional load-to-failure testing on an Instron 5800R machine at a rate of 0.5° per second. Failure was defined as the moment of fracture and subsequent decrease in torsional load—determined graphically by the peak torsional load followed immediately by a sharp decrease in load. Stiffness was determined as the slope of torque to the displacement curve for each cycle, and torque to failure was the highest recorded torque before fracture. Fracture pattern was noted after failure. A sample size of 6 specimens per group provided 80% power to detect a between-group difference of 1 Nm per degree in stiffness, using an estimated SD of 0.7 Nm per degree. In our statistical analysis, continuous variables are reported as means and SDs. Data from our torsional stiffness and load-to-failure testing were analyzed with unpaired 2-sample t tests, and P < .05 was considered statistically significant.

Results

We did not detect a statistical difference in notch depth, notch-to-implant distance, or femoral length between the CR and PS groups. Mean (SD) notch depth was 6.0 (1.3) mm for CR and 4.9 (1.0) mm for PS (P = .13); mean (SD) distance from the proximal end of the notch to the implant was 13.8 (1.7) mm for CR and 11.1 (3.2) mm for PS (P = .08); and mean (SD) femoral length was 46.2 (0.1) cm for CR and 46.2 (0.1) cm for PS (P = .60).

Mean (SD) torsional stiffness for the first 3 precycles was 6.2 (1.2), 8.7 (1.5), and 8.8 (1.4) Nm per degree for the CR group and 6.0 (0.7), 8.4 (1.4), and 8.6 (1.4) Nm per degree for the PS group; the differences were not statistically significant (Figure 2A). In addition, there were no statistically significant differences in mean (SD) stiffness at failure between CR, 6.5 (0.7) Nm per degree, and PS, 7.1 (0.9) Nm per degree (P = .24; Figure 2B) or in mean (SD) final torque at failure between CR, 62.4 (9.4) Nm, and PS, 62.7 (12.2) Nm (P = .95; Figure 2C).

All fractures in both groups were oblique fractures originating at the proximal angle of the notch and extended proximally. None extended distally into the box. Fracture locations and patterns were identical in the CR and PS groups of femurs (Figure 3).

Discussion

Periprosthetic fractures after TKA remain rare. However, these fractures can significantly increase morbidity and complications. Anterior femoral notching occurs inadvertently in 30% to 40% of TKAs.^6,17 The impact of femoral notching on supracondylar femur fracture is inconsistent between biomechanical and retrospective clinical studies. Retrospective studies failed to find a significant correlation between anterior femoral notching and supracondylar femur fractures.^6,17 However, findings of biomechanical studies have suggested that a notch 3 mm deep will reduce the torsional strength of the femur by 29%.⁷ Another study, using 3-dimensional finite element analysis, showed a significant increase in local stress with a notch deeper than 3 mm.¹⁵

To our knowledge, no clinical studies, including the aforementioned Cochrane review,¹⁰ have specifically evaluated the difference in risk for periprosthetic fracture between different TKA models in the presence of notching.¹¹ The biomechanical differences between implant designs could be a confounding factor in the results of past studies. More bone resection is required in PS designs than in CR designs. The position of the PS intercondylar cutout, much lower than the top of the patella flange, should not increase susceptibility to fractures more than in CR designs, but this hypothesis, though accepted, has not been validated biomechanically or addressed specifically in prospective or retrospective clinical analysis. In the present study, we used a biomechanical model to replicate an external rotation failure mechanism and quantify the differences in torsional strength and load to failure between CR TKA and PS TKA models in the presence of anterior femoral notching. Our results showed no significant differences in torsional stiffness, stiffness at failure, or torque at failure between the CR and PS design groups in the presence of anterior femoral notching.

In this study, all femoral fractures were oblique, and they all originated at the site of the cortical defect, not the notch—a situation markedly different from having bending forces applied to the femur. Previous biomechanical data indicated that bending forces applied to a notched femur cause fractures originating at the notch, whereas torsional forces applied to a notched femur cause fractures originating at the anterior aspect of the bone–component interface.⁷ The difference is attributable to study design. Our femurs were held fixed at their proximal end, which may have exacerbated any bending forces applied during external rotation, but we thought constraining the proximal femur would better replicate a fall involving external rotation.

More important for our study, an oblique fracture pattern was noted for both design groups (CR and PS), indicating the fracture pattern was unrelated to the area from which bone was resected for the PS design. All femur fractures in both design groups occurred proximal to a well-fixed prosthesis, indicating they should be classified as Vancouver C fractures. This is significant because intercondylar fossa resection (PS group) did not convert the fractures into Vancouver B2 fractures, which involve prosthesis loosening caused by pericomponent fracture.¹⁸ This simple observation validated our hypothesis that there would be no biomechanical differences between CR and PS designs with respect to the effects of anterior femoral notching. This lack of a significant difference may be attributed to the PS intercondylar cutout being much lower than the top of the anterior flange shielding the resected bone deep to the anterior flange.⁷ In addition, given the rarity of supracondylar fractures and the lack of sufficient relevant clinical data, it is difficult to speculate on the fracture patterns observed in clinical cases versus biomechanical studies.¹²

The use of synthetic bone models instead of cadaveric specimens could be seen as a limitation. Although synthetic bones may not reproduce the mechanism of failure in living and cadaveric femurs, the mechanical properties of synthetic bones have previously been found to fall within the range of those of cadaveric bones under axial loading, bending, and torsion testing.^13,14 As a uniform testing material, synthetic bones allow removal of the confounding variations in bone size and quality that plague biomechanical studies in cadaveric bones.^13,14 Interfemoral variability was 20 to 200 times higher in cadaveric femurs than in synthetic bones, which makes synthetic femurs preferable to cadaveric femurs, especially in studies with a small sample size.^13,14 In addition, a uniform specimen provides consistent, reproducible osteotomies, which were crucial for consistent mechanical evaluation of each configuration in this study.

The long-term clinical significance of anterior femoral notching in periprosthetic fractures is equivocal, possibly because most studies predominantly use CR implants.⁶ This may not be an issue if it is shown that CR and PS implants have the same mechanical properties. Despite the differences between clinical studies and our biomechanical study, reevaluation of clinical data is not warranted given the biomechanical data we present here. Results of biomechanical studies like ours still suggest an increased immediate postoperative risk for supracondylar fracture after anterior cortical notching of the femur.^5,7 Ultimately, this study found that, compared with a CR design, a PS design did not alter the torsional biomechanical properties or fracture pattern of an anteriorly notched femur.

Although rare, periprosthetic fractures remain a significant complication after total knee arthroplasty (TKA), occurring in 0.3% to 2.5% of cases.^1-4 Hirsh and colleagues⁵ were among the first to suggest that anterior femoral notching during TKA was a potential risk factor for postoperative periprosthetic femoral fracture because notching may weaken the anterior femoral cortex. Anterior femoral notching, a cortex violation occurring during an anterior bone cut, occurs in up to 30% of cases.⁶ Using a theoretical biomechanical model, Culp and colleagues¹ found that increasing the depth of the notch defect into the cortex led to reduced torsional strength. In more recent, cadaveric biomechanical studies, notching of the anterior femoral cortex decreased torsional strength by up to 39%.^7,8 Contrary to these biomechanical studies, a retrospective study evaluating 1089 TKAs using 2 implant designs (Anatomic Graduated Component, Biomet and Legacy, Zimmer) demonstrated no significant effect of anterior femoral notching with respect to incidence of supracondylar femur fractures.⁶ That study, however, did not address whether implant design is associated with a differential risk for fracture in the presence of anterior notching.

Previous biomechanical studies have primarily investigated cruciate-retaining (CR) femoral components and properties with respect to anterior notching, even though the posterior-stabilized (PS) design is used more often in the United States.^1,7 According to a Mayo Clinic survey, TKAs with a PS design increased from <10% in 1990 to almost 75% by 1997.⁹ Today, there is little to no consensus about which implant is better, and use of one or the other depends largely on the surgeon and varies widely between countries and regions.¹⁰ PS designs require more bone resection and demonstrate prosthesis-controlled rollback during flexion, whereas CR designs preserve more bone and achieve posterior stabilization via the posterior cruciate ligament.¹¹ Despite these differences in design and mechanics, a 2013 Cochrane review of TKA design found no clinically significant differences between CR and PS with respect to pain, range of motion, or clinical and radiologic outcomes.¹⁰ The reviewers did not specifically address periprosthetic fractures associated with either femoral notching or TKA design, as they could not quantitatively analyze postoperative complications because of the diversity of reports. Given the limited number of reported cases, a review of radiographic findings pertaining to the characteristics of supracondylar fractures in anterior femoral notching was unsuccessful.¹² As the previous biomechanical studies of anterior notching used primarily CR models or no prostheses at all, a study of biomechanical differences between CR and PS designs in the presence of anterior notching is warranted.^1,7,8 Therefore, we conducted a study to assess the effect of anterior femoral notching on torsional strength and load to failure in CR and PS femoral components.

Materials and Methods

Twelve fourth-generation composite adult left femur synthetic sawbones (Sawbones; Pacific Research Laboratories) were selected for their consistent biomechanical properties, vs those of cadaveric specimens; in addition, low intersample variability made them preferable to cadaveric bones given the small sample used in this study.^13,14 All bones were from the same lot. All were visually inspected for defects and found to be acceptable. In each sample, an anterior cortical defect was created by making an anterior cut with an undersized (size 4) posterior referencing guide. In addition, the distance from the proximal end of the notch to the implant fell within 15 mm, as that is the maximum distance from the implant a notch can be placed using a standard femoral cutting jig.¹⁵ Six femora were instrumented with CR implants and 6 with PS implants (DePuy Synthes). Implants were placed using standardized cuts. Before testing, each implant was inspected for proper fit and found to be securely fastened to the femur. In addition, precision calipers were used to measure notch depth and distance from notch to implant before loading. A custom polymethylmethacrylate torsion jig was used to fix each instrumented femur proximally and distally on the femoral implant (Figure 1). Care was taken to ensure the distal jig engaged only the implant, thus isolating the notch as a stress riser. Each femur was loaded in external rotation through the proximal femoral jig along the anatomical axis. Use of external rotation was based on study findings implicating external rotation of the tibia as the most likely mechanism for generating a fracture in the event of a fall.¹² Furthermore, distal femur fractures are predominantly spiral as opposed to butterfly or bending—an indication that torsion is the most likely mechanism of failure.¹⁶ With no axial rotation possible within the prosthesis, increased torsional stress is undoubtedly generated within adjacent bone. Each specimen underwent torsional stiffness testing and then load to failure. Torsional stiffness was measured by slowly loading each femur in external rotation, from 1 to 18 Nm for 3 cycles at a displacement rate of 0.5° per second. Each specimen then underwent torsional load-to-failure testing on an Instron 5800R machine at a rate of 0.5° per second. Failure was defined as the moment of fracture and subsequent decrease in torsional load—determined graphically by the peak torsional load followed immediately by a sharp decrease in load. Stiffness was determined as the slope of torque to the displacement curve for each cycle, and torque to failure was the highest recorded torque before fracture. Fracture pattern was noted after failure. A sample size of 6 specimens per group provided 80% power to detect a between-group difference of 1 Nm per degree in stiffness, using an estimated SD of 0.7 Nm per degree. In our statistical analysis, continuous variables are reported as means and SDs. Data from our torsional stiffness and load-to-failure testing were analyzed with unpaired 2-sample t tests, and P < .05 was considered statistically significant.

Results

We did not detect a statistical difference in notch depth, notch-to-implant distance, or femoral length between the CR and PS groups. Mean (SD) notch depth was 6.0 (1.3) mm for CR and 4.9 (1.0) mm for PS (P = .13); mean (SD) distance from the proximal end of the notch to the implant was 13.8 (1.7) mm for CR and 11.1 (3.2) mm for PS (P = .08); and mean (SD) femoral length was 46.2 (0.1) cm for CR and 46.2 (0.1) cm for PS (P = .60).

Mean (SD) torsional stiffness for the first 3 precycles was 6.2 (1.2), 8.7 (1.5), and 8.8 (1.4) Nm per degree for the CR group and 6.0 (0.7), 8.4 (1.4), and 8.6 (1.4) Nm per degree for the PS group; the differences were not statistically significant (Figure 2A). In addition, there were no statistically significant differences in mean (SD) stiffness at failure between CR, 6.5 (0.7) Nm per degree, and PS, 7.1 (0.9) Nm per degree (P = .24; Figure 2B) or in mean (SD) final torque at failure between CR, 62.4 (9.4) Nm, and PS, 62.7 (12.2) Nm (P = .95; Figure 2C).

All fractures in both groups were oblique fractures originating at the proximal angle of the notch and extended proximally. None extended distally into the box. Fracture locations and patterns were identical in the CR and PS groups of femurs (Figure 3).

Discussion

Periprosthetic fractures after TKA remain rare. However, these fractures can significantly increase morbidity and complications. Anterior femoral notching occurs inadvertently in 30% to 40% of TKAs.^6,17 The impact of femoral notching on supracondylar femur fracture is inconsistent between biomechanical and retrospective clinical studies. Retrospective studies failed to find a significant correlation between anterior femoral notching and supracondylar femur fractures.^6,17 However, findings of biomechanical studies have suggested that a notch 3 mm deep will reduce the torsional strength of the femur by 29%.⁷ Another study, using 3-dimensional finite element analysis, showed a significant increase in local stress with a notch deeper than 3 mm.¹⁵

To our knowledge, no clinical studies, including the aforementioned Cochrane review,¹⁰ have specifically evaluated the difference in risk for periprosthetic fracture between different TKA models in the presence of notching.¹¹ The biomechanical differences between implant designs could be a confounding factor in the results of past studies. More bone resection is required in PS designs than in CR designs. The position of the PS intercondylar cutout, much lower than the top of the patella flange, should not increase susceptibility to fractures more than in CR designs, but this hypothesis, though accepted, has not been validated biomechanically or addressed specifically in prospective or retrospective clinical analysis. In the present study, we used a biomechanical model to replicate an external rotation failure mechanism and quantify the differences in torsional strength and load to failure between CR TKA and PS TKA models in the presence of anterior femoral notching. Our results showed no significant differences in torsional stiffness, stiffness at failure, or torque at failure between the CR and PS design groups in the presence of anterior femoral notching.

In this study, all femoral fractures were oblique, and they all originated at the site of the cortical defect, not the notch—a situation markedly different from having bending forces applied to the femur. Previous biomechanical data indicated that bending forces applied to a notched femur cause fractures originating at the notch, whereas torsional forces applied to a notched femur cause fractures originating at the anterior aspect of the bone–component interface.⁷ The difference is attributable to study design. Our femurs were held fixed at their proximal end, which may have exacerbated any bending forces applied during external rotation, but we thought constraining the proximal femur would better replicate a fall involving external rotation.

More important for our study, an oblique fracture pattern was noted for both design groups (CR and PS), indicating the fracture pattern was unrelated to the area from which bone was resected for the PS design. All femur fractures in both design groups occurred proximal to a well-fixed prosthesis, indicating they should be classified as Vancouver C fractures. This is significant because intercondylar fossa resection (PS group) did not convert the fractures into Vancouver B2 fractures, which involve prosthesis loosening caused by pericomponent fracture.¹⁸ This simple observation validated our hypothesis that there would be no biomechanical differences between CR and PS designs with respect to the effects of anterior femoral notching. This lack of a significant difference may be attributed to the PS intercondylar cutout being much lower than the top of the anterior flange shielding the resected bone deep to the anterior flange.⁷ In addition, given the rarity of supracondylar fractures and the lack of sufficient relevant clinical data, it is difficult to speculate on the fracture patterns observed in clinical cases versus biomechanical studies.¹²

The use of synthetic bone models instead of cadaveric specimens could be seen as a limitation. Although synthetic bones may not reproduce the mechanism of failure in living and cadaveric femurs, the mechanical properties of synthetic bones have previously been found to fall within the range of those of cadaveric bones under axial loading, bending, and torsion testing.^13,14 As a uniform testing material, synthetic bones allow removal of the confounding variations in bone size and quality that plague biomechanical studies in cadaveric bones.^13,14 Interfemoral variability was 20 to 200 times higher in cadaveric femurs than in synthetic bones, which makes synthetic femurs preferable to cadaveric femurs, especially in studies with a small sample size.^13,14 In addition, a uniform specimen provides consistent, reproducible osteotomies, which were crucial for consistent mechanical evaluation of each configuration in this study.

The long-term clinical significance of anterior femoral notching in periprosthetic fractures is equivocal, possibly because most studies predominantly use CR implants.⁶ This may not be an issue if it is shown that CR and PS implants have the same mechanical properties. Despite the differences between clinical studies and our biomechanical study, reevaluation of clinical data is not warranted given the biomechanical data we present here. Results of biomechanical studies like ours still suggest an increased immediate postoperative risk for supracondylar fracture after anterior cortical notching of the femur.^5,7 Ultimately, this study found that, compared with a CR design, a PS design did not alter the torsional biomechanical properties or fracture pattern of an anteriorly notched femur.