Original article
Self-reports of health care utilization compared to provider records

https://doi.org/10.1016/S0895-4356(00)00261-4Get rights and content

Abstract

This study compares self-reports of medical utilization with provider records. As part of a chronic disease self-management intervention study, patients completed self-reports of their last six months of health care utilization. A subgroup of patients was selected from the larger study and their self-reports of utilization were compared to computerized utilization records. Consistent with earlier studies, patients tended to report less physician utilization than was recorded in the computerized provider records. However, they also tended to report slightly more emergency room visits than were reported in the computerized utilization records. There was no association between demographic or health variables and the tendency toward discrepancy between self-report and computerized utilization record reports. However, there was a tendency for the discrepancy to increase as the amount of record utilization increased. Thus, the likelihood of bias caused by differing demographic factors is low, but researchers should take into account that underreporting occurs and is likely to increase as utilization increases.

Introduction

Health care utilization is a common outcome measure in health intervention and other studies [1]. The frequency of visits to physicians may be an indicator both of a patient's health and of a patient's ability to self-manage his or her condition. Utilization may be measured through chart audits, computer record summaries, or self-report, but chart audit is usually considered the “gold standard” for this measurement.

There are several potential problems with using chart audits to measure health care utilization. First, chart audits are cumbersome and costly, especially in large studies. Second, some patients utilize multiple service providers, necessitating accessing multiple charts for each study participant. Third, some visits may not be recorded in the chart, especially visits with health professionals other than physicians. Fourth, some chart entries may be difficult to read or decipher. And fifth, for privacy or security reasons some health care providers may restrict access to charts even with patient consent.

With the use of computerized databases to maintain clinical records becoming common, summaries of utilization can often be produced by computer, generally at lower cost than chart audits. Some of the difficulties of chart audits may be remedied with the use of computerized medical utilization records, but new problems may be introduced. Some computerized utilization records are designed to track patient billings rather than to follow medical histories and may prove inaccurate for measuring specific kinds of utilization. Privacy and security concerns may even increase in terms of allowing researchers access to computerized databases. Furthermore, computerized utilization records may contain incomplete cross-references to care outside the primary provider system, thus making the problem of accessing multiple sources, often with incompatible data systems, even more complex.

In contrast, self-reports are less expensive to obtain and are likely to be inclusive of all sources of health care, but serious questions arise as to whether patient self-reports of utilization are accurate. Several studies have found that patients tend to underreport their utilization of health care services relative to provider records 2, 3, 4, 5, 6. For example, in a study by Yafee and colleagues, only 72.9% of physician office visits found in record checks were reported by the participants [5].

Reporting error may also vary by patient characteristics such as age, gender, education, income, and health status of the patient. Thus, systematic biases could be introduced into studies that compare groups that differ on such characteristics. Although the associations between the reporting error and these factors have been investigated in several studies, the results have been inconsistent. Jobe and colleagues found no significant associations between self-report versus medical record discrepancy scores and gender, age, education or health status [2]. However, in a study by Cleary et al. [3], being older and having lower income were significantly and positively associated with reporting error. We might also expect reporting error to increase with greater utilization. Roberts and colleagues found an increasing bias toward underreporting as the number of ambulatory visits increased in the previous year [6].

The purpose of this article is to compare self-reports to computerized utilization records of three types of utilization and to examine the nature of any discrepancies. We believe that this is one of the first comparisons of self-reports with computerized utilization records. In particular, we are interested in whether discrepancies reflect random error in reporting or whether they might introduce biases into studies that rely on self-reports. If self-report can be shown to be sufficiently accurate and relatively unaffected by demographic and other variables, then self-reports could be considered a viable method for assessing health care utilization in patient education and other intervention studies.

Section snippets

Participants

The data presented in this article came from participants in a larger study of the effectiveness of a Chronic Disease Self-Management Program (CDSMP). All of the participants in the CDSMP had at least one symptomatic disease (either heart disease, lung disease, arthritis, or stroke), were 40 years of age or older, and had volunteered to take a 7-week chronic disease self-management course. Participants were recruited in a number of ways, including media announcements and through doctor

Characteristics of the samples

Table 1 describes the subjects demographic characteristics and the self-reported utilization scores. The sample can be characterized as consisting of older and well-educated individuals, with an average of two chronic conditions.

Magnitude and direction of visits to physicians

The top row of Table 2 compares self-reports to computerized utilization records for physician visits. We found an average Total Discrepancy of 1.06 fewer visits reported per person than were found in the computer records. These numbers represent the cumulative effect

Discussion

This is one of the first comparisons of self-reports of medical utilization with a provider computerized utilization record. The discrepancy rate for MD visits in this study is slightly higher than typically found in other studies of older persons, using medical chart records. For example, Glandon et al. [4] found a total discrepancy for the number of MD visits of −.35 (labeled “total error”) compared to −1.06 in this study. This may be due to the higher levels of utilization in our chronically

Acknowledgements

This study was supported by University of California Tobacco-Related Disease Research Program grant TR156 and AHCPR grant 5RO1 HS06680

References (9)

There are more references available in the full text version of this article.

Cited by (346)

View all citing articles on Scopus
View full text