Agreement Between Self-Report and Medical Record Prevalence of 16 Chronic Conditions in the Alaska EARTH Study

Abstract

The gold standard for health information is the health record. Hospitalization and outpatient diagnoses provide health systems with data on which to project health costs and plan programmatic changes. Although health record information may be reliable and perceived as accurate, it may not include population-specific information and may exclude care provided outside a specific health care facility. Sole reliance on medical record information may lead to underutilization of health care services and inadequate assessment of population health status. In this study, we analyzed agreement, without assuming a gold standard, between self-reported and recorded chronic conditions in an American Indian/Alaska Native cohort. Self-reported health history was collected from 3821 adult participants of the Alaska EARTH study during 2004-2006. Participant medical records were electronically accessed and reviewed. Self-reported chronic conditions were underreported in relation to the medical record and both information sources reported the absence more reliably than the presence of conditions (across conditions, median positive predictive value = 64%, median negative predictive value = 94%). Agreement was affected by age, gender, and education. Differences between participant- and provider-based prevalence of chronic conditions demonstrate why health care administrators and policy makers should not rely exclusively on medical record–based administrative data for a comprehensive evaluation of population health.

Identification of diabetes, heart disease, hypertension and stroke in mid‐ and older‐aged women: Comparing self‐report and administrative hospital data records

“…Good agreement for diabetes is consistent with other literature, and the view expressed that familiar diagnoses associated with conditions that involve frequent interactions with the healthcare system are generally accurately reported in surveys and also more likely to be recorded as comorbidities in hospital records. 3,15,16,18,20 Substantial to excellent agreement for diabetes has been reported by Lix et al (κ = 0.65-0.83), 16 Muggah et al 18 in hospital records if it is not relevant to that admission. 18 Differences found in the ascertainment of heart disease could have been a result of different interpretations of the illness.…”

“…(κ = 0.66–0.80), Koller et al . (κ = 0.6–0.8), Okura et al . (κ = 0.71–0.80), and by Simpson et al .…”

“…A study by Lix et al . found fair to moderate agreement for heart disease among men and women (κ = 0.29–0.55), 46% of whom were aged 50 years and over; whereas Koller found only poor to fair agreement for heart disease among persons aged 65 years and older (κ = 0.34) …”

“…The aims of the present study were: to estimate the prevalence of diabetes, heart disease, hypertension and stroke in self‐report and administrative hospital data records in two cohorts of mid‐ and older‐aged Australian women; to measure sensitivity and agreement between data sources; and to compare the results between age cohorts. As in other agreement studies that compare health administrative data with patient self‐report, hospital records are used here as a point of reference rather than a “gold standard” …”

“…As in other agreement studies that compare health administrative data with patient selfreport, hospital records are used here as a point of reference rather than a "gold standard". 16,18,20…”