Introduction
The proportion of the population that is older is rapidly increasing, and life expectancy is growing in most western countries. Concurrently, the percentage of chronically ill and disabled older adults among the older population is also growing, particularly among the very old. Older people can live longer with chronic diseases if they are properly treated and monitored. One precondition for this is that older people are knowledgeable about their health conditions. It is therefore important to study the extent to which there is concordance between self-reports of chronic health conditions and physician reports or medical records from two perspectives. First, from a research perspective, the study can provide an estimate of the validity, reliability, and usefulness of current self-reporting measures for research purposes and for estimating the presence of chronic diseases among the older population, especially because medical record review is costly (Tisnado et al., Reference Tisnado, Adams, Liu, Damberg, Chen and Hu2006). Second, from a clinical practice perspective, the extent to which there is discordance between these two sources of information may reflect lack of or inaccurate knowledge or interpretation and perceptions, recall problems, and attitudes of the respondents about their health status.
Likewise, there may be non-reporting or misreporting of diagnoses in medical records (Bush, Miller, Golden, & Hale, Reference Bush, Miller, Golden and Hale1989). Discordance can result in treatment deficiency that may consequently hinder the well-being of older people who are chronically ill and functionally disabled. For example, Hong, Oddone, Dudley, and Bosworth (Reference Hong, Oddone, Dudley and Bosworth2005) found that the degree of concordance had implications for the extent to which patients had difficulties managing their illness. Raising awareness among health providers and care recipients and their families in detecting discordance can help to reduce it and increase the likelihood of appropriate treatment.
There is reason to examine discordance with regard to older adults who are functionally disabled as a result of various chronic diseases. Research results have been found to vary by study population, diagnosis, and study design (Gupta et al., Reference Gupta, Gu, Chen, Lu, Shu and Ying Zheng2011). Moreover, most previous studies have focused on the reliability of one or a few conditions simultaneously (e.g., Huerta, Tormo, Egea-Caparrós, Ortolá-Devesa, & Navarro, Reference Huerta, Tormo, Egea-Caparrós, Juan B Ortolá-Devesa and Navarro2009; Skinner, Miller, Lincoln, Lee, & Kazis, Reference Skinner, Miller, Lincoln, Lee and Kazis2005).
To address these limitations, this study examined 13 self-reported chronic medical conditions, which – when compared with diagnostic codes from available electronic medical records – provided insight into the concordance between these two sources of information in disabled older adults. The goals of this study were twofold: first, to examine the extent of concordance between self-reports and medical records on diagnosed chronic health conditions among physically disabled older adults; and second, to identify socio-demographic and health characteristics that are connected with concordance between these two sources of information, such as age, gender, education, and ethnicity. We hypothesized that there would be concordance between these two sources of reports because chronically ill older patients are likely to be in the medical care system for longer periods and therefore may be more aware of their health status, leading to more accurate self-reports of their health conditions. However, degree of concordance may also depend on the type of health condition a person has, taking into account that self-reported health conditions may differ from medical record diagnoses for various reasons, including poor patient-clinician communication, self-diagnosis in the absence of a satisfactory explanation for symptoms, lack of awareness among patients, or the perceptions of patients. As well, concordance may depend on individuals’ socio-demographic characteristics, their functional status, and co-morbidity.
Review of the Literature
Concordance between Self-Reported Health Conditions and Medical Records
In general, research findings indicate good concordance between self-reports and medical records of older people, although concordance varies by medical condition (Bush et al., Reference Bush, Miller, Golden and Hale1989; Corser et al., Reference Corser, Sikorskii, Olumo, Stommel, Proden and Holmes-Rovner2008; Osler & Schroll, Reference Osler and Schroll1992; St. Sauver et al., Reference St. Sauver, Hagen, Cha, Bagniewski, Mandrekar and Curoe2005; Tisnado et al., Reference Tisnado, Adams, Liu, Damberg, Chen and Hu2006). Skinner et al. (Reference Skinner, Miller, Lincoln, Lee and Kazis2005), who examined concordance between self-reports and medical records for older patients having five chronic health conditions (hypertension, diabetes mellitus, chronic low-back pain, osteoarthritis of the knee, and chronic lung disease), found that discordance was smaller for diabetes and hypertension but larger for osteoarthritis of the knee and chronic low-back pain. Simpson et al. (Reference Simpson, Boyd, Carlson, Griswold, Guralnik and Fried2004), who examined concordance in disabled older women with chronic diseases, found high concordance for hip fracture, Parkinson’s disease, diabetes, cancer, stroke, and disc disease; fair to good concordance for angina pectoris, congestive heart failure, and myocardial infarction; and poor concordance for peripheral arterial disease, spinal stenosis, osteoporosis, arthritis, and lung disease. The discordance was in the direction of patients reporting medical conditions that were not documented in their medical charts.
Internationally, Netherlands researchers Kriegsman, Penninx, van Eijk, Boeke, and Deeg (Reference Kriegsman, Penninx, van Eijk, Boeke and Deeg1996) reported finding kappa values ranging from 0.30 to 0.40 for osteoarthritis/rheumatoid arthritis and atherosclerosis, and to 0.85 for diabetes mellitus. Kjvinen, Halonen, Eronen, and Nissinen (Reference Kjvinen, Halonen, Eronen and Nissinen1998) found significant differences in reports of the prevalence among older Finnish men of coronary heart disease and chronic lung disease between patients and their physicians. In Spain (Huerta et al., Reference Huerta, Tormo, Egea-Caparrós, Juan B Ortolá-Devesa and Navarro2009), concordance was found to be high with regard to diabetes, moderate for hypertension, and minimal for hyperlipidemia. In Israel (Gross, Bentur, Alhayany, Sherf, & Epstein, Reference Gross, Bentur, Alhayany, Sherf and Epstein1996), more than half of the respondents with chronic diseases failed to report at least one disease, and the rate of under-reporting varied by type of disease.
In summary, although substantial concordance was found between self-reported chronic morbidity and medical records, particularly with regard to diabetes, research findings have also been found to vary by type of disease and by country (with different reporting systems, measurement differences, and differences in access to physicians; Gupta et al., Reference Gupta, Gu, Chen, Lu, Shu and Ying Zheng2011). It is, therefore, important to also examine differences in this regard and the implications for health care services in multicultural societies, such as in Israel.
Socio-demographic Characteristics and Concordance between Self-Reports and Medical Records
Results from previous studies show (a) that concordance can vary not only by medical condition but also by patients’ characteristics (Corser et al., Reference Corser, Sikorskii, Olumo, Stommel, Proden and Holmes-Rovner2008) and (b) that socio-demographic characteristics play a role in this regard. For example, concordance has been found to vary by age, gender, education (Okura, Urban, Mahoney, Jacobsen, & Rodeheffer, Reference Okura, Urban, Mahoney, Jacobsen and Rodeheffer2004; St. Sauver et al., Reference St. Sauver, Hagen, Cha, Bagniewski, Mandrekar and Curoe2005), and ethnicity (Natarajan, Lipsitz, & Nietert, Reference Natarajan, Lipsitz and Nietert2002). However, the findings are inconsistent.
For example, Corser et al. (Reference Corser, Sikorskii, Olumo, Stommel, Proden and Holmes-Rovner2008) and St. Sauver et al. (2005) found that concordance between self-reports and medical records increased with age. In contrast, Simpson et al. (Reference Simpson, Boyd, Carlson, Griswold, Guralnik and Fried2004) found that concordance decreased with increasing age. Kriegsman et al. (Reference Kriegsman, Penninx, van Eijk, Boeke and Deeg1996) found both over-reporting and under-reporting of cardiac disease, over-reporting of stroke, and under-reporting of arthritis as people became older. Finally, Bowlin et al. (Reference Bowlin, Morrill, Nafziger, Jenkins, Lewis and Pearson1993) found that increasing age was associated with decreased accuracy of self-reported hypertension and hypercholesterolemia, although other studies (e.g., Bergmann, Byers, Freedman, & Mokdad, Reference Bergmann, Byers, Freedman and Mokdad1998) found no effect of age on concordance between self-reports and medical record health conditions.
Similarly inconsistent findings have been reported with regard to gender. For example, Kriegsman et al. (Reference Kriegsman, Penninx, van Eijk, Boeke and Deeg1996) found that, for chronic non-specific lung disease, under-reporting and over-reporting were more prevalent in males than in females. Men tended to over-report stroke and under-report malignancies and arthritis, whereas women tended to over-report malignancies and arthritis. Both over- and under-reporting of cardiac disease were more prevalent as people aged. Johansson, Hellenius, Elofsson, and Krakau (Reference Johansson, Hellenius, Elofsson and Krakau1999) found that women under-reported hypertension and hyperlipidemia whereas other studies have found no effect of gender on concordance between the two sources of reports (e.g., Bowlin et al., Reference Bowlin, Morrill, Nafziger, Jenkins, Lewis and Pearson1993; Bush et al., Reference Bush, Miller, Golden and Hale1989).
With regard to factors other than age and gender, Simpson et al. (Reference Simpson, Boyd, Carlson, Griswold, Guralnik and Fried2004) found that concordance decreased with decreasing cognition and education, and when the patients had four or more chronic diseases. Kjvinen et al. (Reference Kjvinen, Halonen, Eronen and Nissinen1998) found no differences in marital status between self-rated or physician-rated severity of illnesses. Other findings regarding education levels have been inconsistent. For example, a study conducted in Taiwan (Goldman, Lin, Weinstein, & Lin, Reference Goldman, Lin, Weinstein and Lin2003) found higher concordance among hypertensive older respondents with higher formal education, particularly those with a secondary education, whereas no significant difference in education was found for diabetes. In other studies conducted in the Netherlands and in the United States (Kriegsman et al., Reference Kriegsman, Penninx, van Eijk, Boeke and Deeg1996; Mentz et al., Reference Mentz, Schulz, Mukherjee, Ragunathan, White-Perkins and Israel2012), no significant associations were found between education level and concordance in a variety of chronic health conditions, including hypertension. Still another study (Corser et al., Reference Corser, Sikorskii, Olumo, Stommel, Proden and Holmes-Rovner2008) found that patients’ level of education was inconsistent across different health conditions, with better-educated patients less likely to report congestive heart failure that was not documented in the medical records but more likely to report cancer that was not documented. With regard to ethnicity, Vargas, Burt, Gillum, and Pamuk (Reference Vargas, Burt, Gillum and Pamuk1997) found that concordance in hypertension reports was lower among Mexican men compared to Mexican women, non-Hispanic whites, and non-Hispanic blacks Americans, whereas under-reporting of hypertension was found among non-Hispanic whites compared to Latinos and non-Hispanic blacks (Mentz et al., Reference Mentz, Schulz, Mukherjee, Ragunathan, White-Perkins and Israel2012). In addition, no significant difference in concordance for hypertension was found for marital status (Mentz et al., Reference Mentz, Schulz, Mukherjee, Ragunathan, White-Perkins and Israel2012).
In summary, although various socio-demographic characteristics have been associated with concordance, the findings have been inconsistent. Given the importance of concordance between different sources of information for research and practice, further research is needed on the concordance between self-reports and physician reports. Furthermore, it is projected that the older population, particularly the age group consisting of those aged 85 and older, will significantly increase in the near future. This suggests that more people will suffer from multiple chronic diseases that may not be properly addressed, either because of patients’ lack of knowledge or because of physicians’ errors in diagnoses that may result in inappropriate treatment. Thus, it is important to further investigate concordance for a large group of chronic health conditions among disabled older adults and, also, to examine the extent to which several socio-demographic characteristics, and health and functional status, of older persons play a role in this regard.
Health Care Services in Israel
Concordance between different sources of information is an important issue to the health care system and to health policy makers. In Israel, as in several other countries, including Canada, every citizen is entitled to receive universal health services, free of charge. Family physicians make up the backbone of primary health care, and referrals to specialists or hospitals are made by them. Health service suppliers are paid more per capita for patients who are older, suffer from more chronic and severe illnesses, and live in peripheral regions. Health care services are provided by four non-profit Health Maintenance Organizations (HMOs), of which the Clalit Health Services is the largest in Israel. It provides health care services to more than half of Israel’s population (Bendelak, Reference Bendelak2011).
Methods
Sample
This is a cross-sectional study that included a purposive sample of 402 functionally disabled older adults who were users of adult day-care centers. The respondents were recruited through 11 day-care centers in the southern region of Israel that serve about 1,000 physically disabled older persons. The inclusion criteria were to (a) be over age 60, (b) able to speak Hebrew or Russian (about one third of the respondents were immigrants from former Soviet Union countries and who immigrated to Israel after the collapse of the Soviet regime in 1989), (c) have difficulties performing activities of daily living (ADL), (d) be cognitively intact, and (e) hold membership in the Clalit HMO. The vast majority of older adults in Israel are members of this HMO, and we had access to its computerized database. To ensure variability, all participants in the 11 adult day-care centers who met the criteria were included in the study. In total, 402 persons were interviewed, 165 refused to be interviewed, 75 were not members of the Clalit HMO, and the remainder were unable to be interviewed as a result of language barriers, deafness, or because they were cognitively impaired or unavailable.
Recruitment and Interview Procedure
The study was approved by the ethics committees of the Clalit HMO and the University Medical Center. IRB protocol approval number is 0036-08 (K); informed consent was obtained from all subjects.
In the first stage, a letter was sent to managers of the adult day-care centers explaining the goals of the study and asking for their permission for the researchers to present the research goals to the users of this service. In the second stage, visitors to the adult day-care centers were approached by interviewers and asked to volunteer to be interviewed. Those respondents who gave their consent and who were Clalit HMO members underwent a short mini-mental test (MMSE) (Folstein, Folstein, & McHugh, Reference Folstein, Folstein and McHugh1975) that included 13 questions related to memory and orientation to make sure that they were not cognitively impaired. If respondents gave three or more incorrect answers to questions (e.g., date of birth, age, what day is it today, who is the current prime minister, who was the former prime minister), the interview was stopped. After completing the MMSE, the research goals were explained, and if respondents agreed to be interviewed, they were asked to sign a consent form. Interviews were conducted in one of the rooms at the day-care centers to ensure confidentiality.
Data were collected, using a structured questionnaire, during 2009–2010 through face-to-face interviews. Interviews lasted from 30 to 45 minutes. There were seven experienced interviewers who were proficient in both Hebrew and Russian and who had worked either in health care services or in services for older people. Interviewers underwent a short training program and were provided with written guidelines on how to complete the questionnaire. The questionnaire had three parts: the first addressed the respondents’ health and functional status and details on their chronic diseases; the second related to their use of services and well-being, and the last part of the questionnaire included personal information (e.g., socio-demographic characteristics). After completing the interviews, the researchers retrieved medical information was retrieved from the computerized database of the Clalit HMO. As patients are seen by their family physicians at the community health clinics, diagnoses are recorded on a master diagnostic table in patients’ medical records. The diagnoses are coded using the World Health Organization’s International Classification of Diseases Version 9, Clinical Modification (ICD-9-CM), and are available for computerized retrieval.
Measures
Self-Reported Chronic Health Conditions
Drawing from the Cross-sectional and Longitudinal Aging Study (CALAS; Modan et al., Reference Modan, Fuchs, Blumstien, Chetrit, Lusky and Novikov2002), co-morbidity was assessed using a list of chronic diseases. Specifically, respondents were asked, “Do you suffer from or has a physician ever told you that you suffer from …?” A list of 13 major chronic health conditions was then read out. These conditions included hypertension, diabetes, myocardial infarction, other heart diseases, circulatory disease, respiratory disease, cardiovascular accident (CVA), gastrointestinal disease, osteoporosis, cancer, thyroid disease, arthritis, and renal problems. Yes (= 1) or no (= 0) response choices were provided for each condition. Scores were summed with higher scores reflecting more co-morbidity.
Chronic Health Conditions Based on Medical Records
Chronic health conditions drawn from the computerized medical records of the respondents were coded using the International Classification of Diseases (ICD-9-CM) (WHO, 2007). Comparisons were made for the 13 chronic illnesses that were reported by the respondents. Thus, for each of the 13 health conditions, the specific diagnostic codes were examined (e.g., thyroid disease, diabetes); if the health conditions were too general (e.g., cancer, heart diseases), then categories of diagnostic codes were used, as presented in Table 1. Each condition was scored as yes (=1) and no (=0). Scores were then summed with higher scores reflecting greater co-morbidity.
Table 1: Diagnostic codes
ICD-9-CM = International Classification of Diseases, Version 9, Clinical Modification
Concordance in Co-morbidity
Concordance was defined as being evident when the self-report of a chronic health condition was identical to that drawn from the computerized medical record and was coded 0 = concordance (data from both sources were either positive or negative). Discordance was defined as being evident when the assessments were not identical (and was coded as 1); that is, when chronic health conditions documented in the medical records were not reported by respondents or when self-reported chronic health conditions were not reported in the medical records.
Instrumental Activities of Daily Living (IADL)
Fillenbaum’s (Reference Fillenbaum1985) measure was used to assess the ability to perform instrumental activities of daily living (IADL). The measure includes eight items relating to light, and heavy, home chores; laundry; cooking; shopping; making arrangements; using a phone; and ability to go to the bus station. Respondents were asked to appraise their difficulty to perform each of the eight IADLs. Scores for each item ranged from 1 (no difficulty at all) to 5 (much difficulty). The final index was based on a sum of the scores, ranging from 8 to 40. The internal consistency (Cronbach’s alpha) of this measure in the present study was .90.
Activities of Daily Living (ADL)
ADL ability was measured using Katz, Downs, Cash, and Grotz’s (Reference Katz, Downs, Cash and Grotz1970) index that includes eight items (i.e., washing, dressing, toileting, indoor mobility, eating, combing/teeth brushing, put on shoes/nail cutting, and transferring from bed to chair), with scores for each item ranging from 1 (no difficulty at all) to 5 (much difficulty). Respondents were asked to appraise their difficulty to perform each of the eight ADLs. The sum of scores produced an index ranging from 8 to 40. The internal consistency (Cronbach’s alpha) of this measure was .93.
Socio-demographic Characteristics
These included age, gender, education (seven categories ranging from 1 = partial elementary school to 7 = graduate degree and over), ethnicity (coded as 1 = Asia/Africa and 0 = otherwise), marital status (coded 1 = married and 0 = unmarried), and length of time living in Israel.
Statistical Analyses
A range of descriptive analyses (percentages, means, and standard deviations) were initially performed to present the characteristics of the respondents and distribution of chronic health conditions by source of information. Classification of the health conditions in the medical records to be compared with the self-reported health conditions was performed by one of the researchers who is a professor in family medicine and is familiar with the ICD-9-CM. A chi-square test was carried out to examine the associations between concordance of the two data sources and Cohen’s kappa values for agreement between the two data sources. Binary logistic regression analyses were conducted to examine the factors that explain concordance for each of the medical conditions. The equation included socio-demographic characteristics of the respondents such as age, gender, ethnicity, education, and marital status. In addition, functional status (regarding ADL and IADL) and co-morbidity based on medical records were included in the equations. Data storage and analyses were performed using IBM’s SPSS software, version 17.
Results
Participants’ Characteristics
Participants’ characteristics are presented in Table 2. The findings show that the vast majority (74.8%) were women. The average age of participants was 78.03 years (SD = 7.02). Most participants (72.7%) were widowed, 20.9 per cent were married, and the remainder were divorced or separated. Only 1.7 per cent were born in Israel; 62.6 per cent were born in Asian/African countries. On average, the sample’s level of education was relatively low, with 68.8 per cent having from zero to eight years of education, 21.1 per cent had from 9 to 12 years of schooling, and only 10.1 per cent had higher education. For those who were born outside of Israel, the average length of residence in Israel was 44.98 years (SD = 17.12). Participants’ functional status scores in terms of ADL ranged from 8 to 32, and the majority (approximately 70%) was mildly dependent. With regard to IADL, the scores ranged from 8 to 38, and the majority (approximately 56%) was moderately dependent.
Table 2: Characteristics of respondents (n = 402)
M = mean
SD = standard deviation
Co-morbidity by Source of Information
The average number of chronic health conditions self-reported by participants was 3.39 (SD = 2.20; range 0 to 11) compared to 4.91 (SD = 2.15; range 0 to 12) reported in the medical records. Thus, slightly more chronic health conditions were evident in the medical records compared to self-reports. The correlation coefficient between self-reports and computerized medical records was significantly positive (r = .45, p < .001).
Concordance between Self-Reports and Medical Records
Table 3 presents the four options with regard to concordance between self-reports and medical records for each of the 13 chronic health conditions. Cohen’s kappa values for concordance between the different sources of information are also reported. The findings indicate that the majority of respondents reported chronic health conditions that were in line with those documented in their individual medical records. The highest levels of concordance were found for diabetes, cancer, thyroid diseases, and CVA. Self-reports corresponded with the computerized medical data in 88.1, 85.6, 82.1, and 81.4 per cent of the cases respectively. The lowest levels of agreement were found for arthritic (51.5%), renal (61.8%), and gastrointestinal (63.2%) conditions. However, discordance between the two sources of reports ranged from about 12 per cent for diabetes to almost 48.5 per cent for arthritis. For example, 17.4 per cent of the respondents reported they suffered from renal problems, although this was not documented in their medical records, and 20.9 per cent of the respondents reported that they did not suffer from any renal problems, while the medical records indicated that they had renal problems. Similar findings were evident for all other chronic health conditions. In particular, there was a significant discrepancy for arthritis whereby 42.3 per cent of the respondents did not report this chronic disease, although it was registered in their medical records.
Table 3: Concordance in reported morbidity by source of information (percentage)
a Yes/Yes and No/No = self-report and medical records are identical
b Yes/No = the respondent reported that s/he suffers from this disease, but it was not documented in the medical records
c No/Yes = the respondent reported not suffering from this disease but it was documented in the medical records
*p < .05, **p < .01, ***p < .001
With the exception of diabetes (kappa = 0.76), the kappa statistics obtained for most health conditions ranged from 0.14 (for arthritis) to 0.41 (for thyroid diseases). These relatively low to fair kappa scores indicate that the level of agreement corrected for chance was generally low. For all chronic diseases except diabetes, the two data sources had significant discordance: the most prevalent pattern of discordance involved respondents who reported that they did not suffer from specific chronic health conditions whereas their medical records provided evidence that they did have these specific chronic health conditions. Does this mean that physicians are medicalizing what patients see as “normal aging”? This may be the case for arthritis, osteoporosis, and perhaps hypertension, conditions that, for some patients, are regarded as the natural result of aging.
To examine the factors that explain concordance for each of the health conditions, binary logistic regression analyses were conducted with concordance for each chronic health condition used as the outcome variable. Each equation included the participants’ socio-demographic characteristics, their functional status, and the number of chronic diseases (co-morbidity) based on the medical records as predictors. In Table 4, only those regression analyses that included two or more significant variables that increased or decreased the likelihood of the outcome variable are presented. The findings show that different socio-demographic characteristics were significant in explaining concordance in various health conditions. For example, ethnicity was found to be significant for concordance in arthritis and cancer: being born in Asian/African countries increased the likelihood of discordance with regard to arthritis but decreased the likelihood of discordance with regard to cancer. Age and gender were found to be significant for concordance in hypertension, with being older increasing the likelihood of discordance whereas being a woman decreased the likelihood of discordance. Education was found to be significant for concordance in gastrointestinal diseases and increased the likelihood of discordance while marital status was found to be significant in explaining concordance in thyroid disease. For all health conditions, registered co-morbidity was a significant factor in explaining the concordance of the various health conditions.
Table 4: Binary logistic regression analyses of factors explaining concordancea in health conditions
a Concordance = 0 and discordance = 1.
b Based on medical records
*p < .05, **p < .01, ***p < .001
ADL = activities of daily living
B = Coefficient for the constant (the “intercept”) in the null model
IADL = instrumental activities of daily living
OR = odds ratio
SE = standard error
For all other health conditions, only one independent variable was found to be significant in explaining the outcome variables: for discordance in diabetes, only functional disability (of ADL) was significant and increased the likelihood of discordance (B = .09, SE = .05, Wald = 3.99, OR = 1.10, p < .05). For heart attacks (B = .40, SE = .06, Wald = 40.95, OR = 1.49, p < .001) and other heart diseases (B = .32, SE = .06, Wald = 30.54, OR = 1.38, p < .001, respectively), only co-morbidity was significant and increased the likelihood of discordance. The same was true for circulatory conditions (B = .16, SE = .05, Wald = 9.13, OR = 1.17, p < .01), respiratory conditions (B = .18, SE = .06, Wald = 9.41, OR = 1.19, p < .01), and CVA (B = .23, SE = .06, Wald = 13.18, OR = 1.26, p < .01). For renal problems only ethnicity was significant, and it decreased the likelihood of discordance (B =–.78, SE = .27, Wald = 8.34, OR = .46, p < .01), suggesting that the likelihood for discordance was higher for respondents who were born in Asian/African countries than for their counterparts. For osteoporosis, only gender was significant in explaining the outcome variable and increased the likelihood of discordance (B = .93, SE = .31, Wald = 9.46, OR = 2.58, p < .001), suggesting that when the respondents were women, the likelihood of discordance for osteoporosis was increased.
Discussion
The main objectives of this study were (1) to examine the extent of concordance evident between self-reported and computerized medical record accounts of chronic health conditions among functionally disabled older adults; and (2) to examine the socio-demographic and health factors that explain concordance. The findings indicated that despite agreement between self-reports and medical records for the majority of participants, there was also substantial discordance between the two sources of reports.
The relatively low level of agreement that was evident after adjustment for chance could be attributed to the lower validity of self-reported data, an interpretation that is consistent with several previous studies (e.g., Bush et al., Reference Bush, Miller, Golden and Hale1989; Osler & Schroll, Reference Osler and Schroll1992; Simpson et al., Reference Simpson, Boyd, Carlson, Griswold, Guralnik and Fried2004; St. Sauver et al., Reference St. Sauver, Hagen, Cha, Bagniewski, Mandrekar and Curoe2005). This lower validity of self-reported data may be particularly true when dealing with functionally disabled older persons. However, caution is warranted in that the sample of respondents did not represent all functionally disabled older persons but, instead, only users of adult day-care centers. It might also be the case that the coding used in the medical records may not have been accurate (e.g., physicians may have chosen incorrect diagnostic codes). Future research will need to address this issue in more detail, unraveling possible mechanisms to explain the level of agreement between self-reported and registered medical data among chronically ill and disabled older persons.
Some lack of agreement between self-reported data and registered data is to be expected for several reasons. First, it may reflect the perceptions of health and illnesses that older people, particularly those who are chronically ill and functionally disabled, have about their conditions. It might be that following a physician’s diagnosis, the individual is actively controlling the disease through medication, diet, exercise, or other healthy lifestyle changes, such that the actual health condition does not indicate an active chronic disease. In other words, if the chronic illness is properly treated and controlled, this circumstance may cause patients to perceive that their diseases are manageable and, consequently, to report that they do not suffer from the specific disease.
A second reason for lack of agreement is that patient reports of past medical conditions may be subject to recall bias or misinterpretation of physician diagnoses, particularly if the patient is not under active medical follow-up or in need of medications on a routine basis (such as for heart attack and renal problems). In addition, there might be other potential problems such as mistakes in medical records (including the exact codes of diagnoses) or the lack of adequate documentation (because accurate records are dependent on patients’ sharing information about symptoms). Thus, relying on medical record data alone could cause physicians to miss chronic health conditions that the patient could have reported (Tisnado et al., Reference Tisnado, Adams, Liu, Damberg, Chen and Hu2006). In addition, shortcomings in questionnaire design or in the completeness of medical records registered by family physicians might yield a lower level of agreement. Therefore, more-precise wording of medical terms may facilitate more-accurate responses from respondents, particularly those with lower health literacy or education.
Third, a lack of agreement between self-reports and medical records might result from older persons’ becoming adapted to their chronic health problems. This may explain why about 22 per cent of respondents who had undergone a heart attack, about 13 per cent with cancer, and about 14 per cent who had experienced a CVA did not report these medical conditions that had developed in the past. For example, McCorkle and Quint-Benoliel (Reference McCorkle and Quint-Benoliel1983) examined two life-threatening diseases – lung cancer and heart attack – and found that respondents reported fewer existential concerns and mood disturbances at the second measurement occasion compared to the first, suggesting that patients can adapt to their situations.
Finally, there are older persons who may be unaware of their chronic diseases, such as hypertension or diabetes, because these diseases are asymptomatic in their early stages, are not severe, and therefore are not treated with medications. This is in line with Chaterrji, Joo, and Lahiri’s (Reference Chaterrji, Joo and Lahiri2010) findings indicating that significant proportions of patients with diagnosed hypertension and diabetes were unaware of these chronic health conditions. Lack of awareness of chronic health conditions may also derive from communication barriers between patients and family physicians, due to lack of language congruence, hearing problems, or misunderstanding, as has been found in previous studies (Okura et al., Reference Okura, Urban, Mahoney, Jacobsen and Rodeheffer2004; Paganini-Hill & Chao, Reference Paganini-Hill and Chao1993; St. Sauver et al., Reference St. Sauver, Hagen, Cha, Bagniewski, Mandrekar and Curoe2005; van Wieringen, Harmsen, & Bruijnzeels, Reference van Wieringen, Harmsen and Bruijnzeels2002).
Further, family physicians are regularly overloaded with work; consequently, they have difficulties devoting adequate time to carefully explaining the medical problems to their older patients at the time of diagnosis. This may cause patients’ perceptions to differ from what was documented in the medical records, or they may self-diagnose in the absence of a satisfactory explanation for their health complaints. It is also possible that some patients actively deny their health conditions.
Multivariate analyses showed that registered co-morbidity was significant in explaining discordance for all of the medical conditions studied. Furthermore, in all cases, except for hypertension, discordance significantly increased with co-morbidity. This might be because those with more co-morbidities are given less attention by their family physicians compared to those with acute health conditions (Adelman, Greene, & Ory, Reference Adelman, Greene and Ory2000). In addition, it might be that chronically ill and functionally disabled older persons are often accompanied by a family member or by a paid carer who takes on the role of advocate for the older person, thus preventing the latter from receiving full information regarding his or her health conditions. This may decrease the patient’s knowledge and awareness of any health conditions, which in turn results in more discordance between self-reports and registered medical data.
The multivariate analyses also showed that except for renal problems, cancer, and arthritis, ethnicity was not a significant factor in explaining discordance. This, too, is in line with previous studies (El Fakiri, Bruijnzeels, & Hoes, Reference El Fakiri, Bruijnzeels and Hoes2007; Reijneveld, Reference Reijneveld2000) that found no statistically significant association between ethnicity and concordance. It might be that despite being born in other countries, the relatively long period of living in Israel acculturated the study participants. However, this issue was not examined in our study and merits further investigation. All other socio-demographic variables were generally found to be not significantly connected with concordance between the two sources of data, except for age and gender that were significant in explaining concordance in hypertension, education that was found to be significant in explain concordance in gastrointestinal diseases, and marital status that was found to be significant in explaining concordance in thyroid disease.
Conclusions
This study revealed that substantial proportions of disabled older adults had self-reports that did not match their medical records, which means low validity of self-reported data. This outcome may reflect a combination of factors dependent on the patients as well as on the documentation accuracy of diagnoses made by physicians. In turn, it may also have negative effects on the well-being and quality of life of chronically ill and disabled older persons: they may not be aware of various health conditions and, consequently, may not receive the appropriate medical treatment and may not comply with medical instructions. This possibility suggests that family physicians should increase their awareness of discordance and its possible implications for their older patients. More time should be devoted to their chronically ill patients to provide them with more detailed information and explanations of their health conditions and to ensure that physicians’ explanations are clear and understood by patients. Communication workshops for family physicians can be helpful in achieving these goals. More research is needed, however, to examine these issues and determine the reasons for these discrepancies.
Limitations
This study has several limitations. First, because the sample covered only one geographic area of Israel, other adults in different regions might present different results in the relationship between self-reported and documented morbidity. Furthermore, the study included respondents who were not randomly selected and were functionally disabled. Another potential limitation of this study is that a significant proportion of respondents were not interviewed due to the inclusion criteria, refusals, communication barriers, or unavailability. Therefore, our results may not be generalizable to respondents with characteristics that differed from those of our study population. Generalization of the findings is thus unwarranted and limited.
In addition, our study did not examine the reasons for discordance. It might be that including more information on the attitudes and perceptions of health and morbidity with respect to the respondents’ mental health, as well as communication and interaction patterns between physicians and their patients, may provide better insights on the issues we examined. Nevertheless, this study contributes to the literature by highlighting a little-examined issue with regard to a specific group of older persons who are functionally disabled.
