Generating EQ-5D-3L health utility scores from the Edinburgh Postnatal Depression Scale: a perinatal mapping study

• Published in: The European journal of health economics 2024-03, Vol.25 (2), p.319-332
• Main Authors: Camacho, Elizabeth M., Shields, Gemma E., Chew-Graham, Carolyn A., Eisner, Emily, Gilbody, Simon, Littlewood, Elizabeth, McMillan, Dean, Watson, Kylie, Fearon, Pasco, Sharp, Deborah J.
• Format: Article
• Language: English
• Subjects: Algorithms; Cohort Studies; Economic Policy; Health Care Management; Health Economics; Health services utilization; Health Status; Humans; Medicine; Medicine & Public Health; Mental health; Observational studies; Original Paper; Pharmacoeconomics and Health Outcomes; Postpartum depression; Psychiatric Status Rating Scales; Public Finance; Public Health; Quality of Life; Surveys and Questionnaires
• ISSN: 1618-7598; 1618-7601
• DOI: 10.1007/s10198-023-01589-4

Background Perinatal depression (PND) describes depression experienced by parents during pregnancy or in the first year after a baby is born. The EQ-5D instrument (a generic measure of health status) is not often collected in perinatal research, however disease-specific measures, such as the Edinburgh Postnatal Depression Scale (EPDS) are widely used. Mapping can be used to estimate generic health utility index values from disease-specific measures like the EPDS. Objective To develop a mapping algorithm to estimate EQ-5D utility index values from the EPDS. Methods Patient-level data from the BaBY PaNDA study (English observational cohort study) provided 1068 observations with paired EPDS and EQ-5D (3-level version; EQ-5D-3L) responses. We compared the performance of six alternative regression model types, each with four specifications of covariates (EPDS score and age: base, squared, and cubed). Model performance (ability to predict utility values) was assessed by ranking mean error, mean absolute error, and root mean square error. Algorithm performance in 3 external datasets was also evaluated. Results There was moderate correlation between EPDS score and utility values (coefficient:  – 0.42). The best performing model type was a two-part model, followed by ordinary least squared. Inclusion of squared and cubed covariates improved model performance. Based on graphs of observed and predicted utility values, the algorithm performed better when utility was above 0.6. Conclusions This direct mapping algorithm allows the estimation of health utility values from EPDS scores. The algorithm has good external validity but is likely to perform better in samples with higher health status.