An improved SDoH measure? – Healthcare Economist

A paper by Saulsberry et al. (2023) argues that the Social Vulnerability Metric (SVM) is an enchancment over earlier social determinants of well being (SDOH) measures such because the Social Vulnerability Index (SVI). SVI makes use of census tract stage knowledge to assemble total neighborhood rankings based mostly on variables included in 4 themes:

• “Socioeconomic Standing” together with percentages beneath poverty, unemployed, earnings, no highschool diploma;
• “Family Composition & Incapacity” together with proportions aged 65 or older, aged 17 or youthful, civilian with a incapacity, single-parent households;
• “Minority Standing & Language” together with proportions minority and speaks English “lower than nicely,”
• “Housing Sort & Transportation” together with proportions multi-unit buildings, cell houses, crowding, no car, and group quarters

I’ve summarized how SVI might be used as a part of a distributional value effectiveness evaluation (DCEA) method right here.

Alternatively, the Social Vulnerability Metric (SVM) proposed by Saulsberry et al. (2023) was constructed utilizing multidimensional Merchandise Response Idea (MIRT), utilizing knowledge from the Agency for Healthcare Research and Quality (AHRQ) SDoH Database. Key variations from SVI.

• Significance relatively than equal weighting. SVM was establishing utilizing the MIRT method (particularly a full-information merchandise bifactor mannequin). As in comparison with a regular latent variable regression, MIRT permits for a number of latent variable constructs. The the bifactor mannequin particularly “imposes restrictions on the standard merchandise issue evaluation by requiring that every merchandise masses on a main dimension (e.g., SDoH) and just one subdomain (e.g., bodily infrastructure).” The coefficients from the MIRT are used to weight variables in SVM. This method differs from CDC’s SVI and measures such because the Space Deprivation Index (ADI), which give equal weight to all variables,
• Unit of geography. SVM was based mostly on ZIP code stage observations whereas SVI historically makes use of census tract knowledge. Nevertheless, AHRQ stated that the their SDoH knowledge base will likely be up to date to report on the
  county, zip code, and census tract block teams sooner or later (and SVM will likely be calculated for every of those).

The variables included in SVM span 5 domains.

• Demographic (e.g., age and race/ethnicity),
• Schooling,
• Financial context (e.g., unemployment fee),
• Bodily infrastructure (e.g., housing and transportation),
• Well being care (e.g., medical insurance protection).

Be aware that race/ethnicity just isn’t included in SVM, partially as a result of this fashion SVM can be utilized to check SDoH throughout race and ethnic teams.

Outcomes

When evaluating SVM vs. SVI, SVM did a greater job of predicting all-cause, age-adjusted mortality (r=0.68 vs. r=0.34). SVM was additionally negatively correlated with receipt of a number of COVID-19 vaccinations (r = −0.68) and completion of full vaccination (r = −0.70) and positively correlated with age-adjusted ED visits for bronchial asthma for people aged 0-18 (r = 0.62) and aged 18 years and older (r = 0.60).

Whereas SVM does seem superior to SVI, it’s considerably extra difficult to create provided that weights usually are not equal, and the coefficients in truth span a number of latent variables.

You may learn the total article here.