UBC Faculty Research and Publications

On the validity of area-based income measures to proxy household income Hanley, Gillian E; Morgan, Steve Apr 10, 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12913_2007_Article_647.pdf [ 472.54kB ]
JSON: 52383-1.0132609.json
JSON-LD: 52383-1.0132609-ld.json
RDF/XML (Pretty): 52383-1.0132609-rdf.xml
RDF/JSON: 52383-1.0132609-rdf.json
Turtle: 52383-1.0132609-turtle.txt
N-Triples: 52383-1.0132609-rdf-ntriples.txt
Original Record: 52383-1.0132609-source.json
Full Text

Full Text

ralssBioMed CentBMC Health Services ResearchOpen AcceResearch articleOn the validity of area-based income measures to proxy household incomeGillian E Hanley*1,2 and Steve Morgan1,2Address: 1Centre for Health Services and Policy Research, University of British Columbia, Vancouver, BC Canada and 2Department of Health Care and Epidemiology, University of British Columbia, Vancouver, BC CanadaEmail: Gillian E Hanley* - ghanley@chspr.ubc.ca; Steve Morgan - morgan@chspr.ubc.ca* Corresponding author    Background: This paper assesses the agreement between household-level income data and anarea-based income measure, and whether or not discrepancies create meaningful differences whenapplied in regression equations estimating total household prescription drug expenditures.Methods: Using administrative data files for the population of BC, Canada, we calculate incomedeciles from both area-based census data and Canada Revenue Agency validated household-leveldata. These deciles are then compared for misclassification. Spearman's correlation, kappacoefficients and weighted kappa coefficients are all calculated. We then assess the validity of usingthe area-based income measure as a proxy for household income in regression equations explainingsocio-economic inequalities in total prescription drug expenditures.Results: The variability between household-level income and area-based income is large. Only 37%of households are classified by area-based measures to be within one decile of the classificationbased on household-level incomes. Statistical evidence of the disagreement between incomemeasures also indicates substantial misclassification, with Spearman's correlations, kappacoefficients and weighted kappa coefficients all indicating little agreement. The regression resultsshow that the size of the coefficients changes considerably when area-based measures are usedinstead of household-level measures, and that use of area-based measures smooths out importantvariation across the income distribution.Conclusion: These results suggest that, in some contexts, the choice of area-based versushousehold-level income can drive conclusions in an important way. Access to reliable household-level income/socio-economic data such as the tax-validated data used in this study wouldunambiguously improve health research and therefore the evidence on which health and socialpolicy would ideally rest.BackgroundMeasures of income are often central to health and healthan enabling factor for access to care[4], or a considerationwhen judging equity of policies and programs[5]. AsPublished: 10 April 2008BMC Health Services Research 2008, 8:79 doi:10.1186/1472-6963-8-79Received: 4 December 2007Accepted: 10 April 2008This article is available from: http://www.biomedcentral.com/1472-6963/8/79© 2008 Hanley and Morgan; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Page 1 of 7(page number not for citation purposes)policy research. Among many potential implications,income can be a non-medical determinant of health [1-3]important as this variable may be, it is often difficult forhealth and health policy researchers to obtain reliable,BMC Health Services Research 2008, 8:79 http://www.biomedcentral.com/1472-6963/8/79individual-level income information for the populationsthey study. In the absence of individual-level income data,investigators often supplement health research datasetswith group-based measures such as area-based averageincome constructed from national census data[6]. Suchmeasures are used as proxy for individual-level incomedata on the assumption that household incomes will bereasonably homogeneous within small enough residen-tial areas. If, however, there is significant heterogeneity inthe areas used, then the aggregate measures can result inecological fallacy–when an association observed betweenvariables at an aggregate level does not represent the asso-ciation that exists at an individual level[7].Prior studies have investigated misclassification ofincome and other socio-economic variables by comparingindividual versus area-level survey responses for smallsamples of the population[8,9] and by comparing survey-based measures for different sized census areas[10,11].Using a unique dataset that contains validated householdincome data for approximately 78% of the population ofBritish Columbia (BC), we investigate the level of misclas-sification that can occur when census-defined, area-basedincome is used as a proxy for an individual's actual house-hold-level income. As the question of most interest toresearchers concerns how well aggregate variables performwhen they are entered in health outcomes equations, wethen assess the sensitivity of the analysis of health relatedinequities in total prescription drug costs to whetherincome is measured as an area-based variable or an house-hold-level variable.MethodsDataOur primary datasets are administrative files for the pro-vincially administered, universal public medical and hos-pital health insurance program, Medical Services Plan(MSP) of BC. This program covers virtually all 4.2 millionresidents of BC, excluding only those residents covered byfederal health insurance programs (collectively about 4%of the population). We restrict our attention to house-holds for which one or more member resided in BC for atleast 275 days per year from 2001 to 2004, inclusive.Household income was obtained from the 2004 registra-tion files for provincially administered, universal publicpharmaceutical insurance program, BC PharmaCare. Inaddition to programs for social assistance recipients andother select populations, BC PharmaCare began offeringincome-based public drug coverage to all residents of theprovince in May 2003. Terms such as deductibles and co-insurance are based on household income, with moregenerous but still income-based coverage offered to seniorHealth obtains net, pre-tax income information from theCanada Revenue Agency. Because of differences in cover-age offered and average needs, 95% of households withone or more senior member were registered for Fair Phar-maCare in 2004 whereas only 73% of non-senior house-holds were registered.The area-based income variables used in this study arebased on linking MSP registry postal codes to averagehousehold income in the area as recorded in the 2001Census. Statistics Canada collates average householdincome and composition for over 7,000 Census Dissemi-nation Areas comprised of 400 to 700 persons. Forresearch purposes, these areas are sorted by income andaggregated into 1,000 strata. Income strata contain anaverage of 1,700 households, with some variation due tovariations in populations by postal code. Both the house-hold level and area-based income variables are based onthe same income concept, gross income prior to anydeductions.Total individual expenditures on prescription drugs wereobtained from BC PharmaNet. BC PharmaNet is anadministrative dataset in which every prescription dis-pensed in the province must be entered by law–it isdesigned to support drug dispensing, drug monitoringand claims processing. These individual expenditureswere aggregated at the household level according to regis-tration files for the MSP program to create a variable indi-cating total household spending on prescription drugs.The research data were extracted for this study from theBritish Columbia Linked Health Database and the BCPharmaNet database with permission of the BC Ministryof Health and the College of Pharmacists of BC. Ethicsapproval was obtained from the Behavioural ResearchEthics Board at the University of British Columbia.Statistical methodsThe household-specific and area-based income measureswere each aggregated into deciles (ordered from lowest tohighest income). We assess the discrepancy between thetwo measures using the CRA validated, household-spe-cific incomes as the standard. We calculated the Spear-man's rank correlations of the various income measures,and both the kappa and weighted kappa to measure thedegree of non-random agreement and partial agreementbetween the measures.We proceed to examine whether the choice of incomemeasure has an impact on how pharmaceutical expendi-tures are distributed by income status. We begin by exam-ining the distribution of prescription drug expendituresPage 2 of 7(page number not for citation purposes)citizens (residents aged 65 and older). For all householdsthat registered to receive coverage, the BC Ministry ofby income deciles, where the deciles are defined accordingto household-level income then according to neighbour-BMC Health Services Research 2008, 8:79 http://www.biomedcentral.com/1472-6963/8/79hood level income. As measurement error is accommo-dated more easily in regression analysis than indescriptive analysis, we also include a series of dummyvariables for both versions of the income variable in anOLS regression in order to determine whether both area-based income and household income generate meaning-fully different results when applied in a research context.We perform regressions of income on total drug expendi-tures with and without covariates controlling for the pres-ence of one or more seniors in the household as well ashousehold size. Through the comparison of coefficientsbetween household-level income variables and area-levelincome variables, one can reach some conclusions aboutthe appropriateness of substituting an area-based measurefor a missing household-level variable in a regressionequation. By including regressions with and without cov-ariates, we can determine whether multivariate modelsinfluence the discrepancy between area-based and house-hold-level variables.ResultsA total of 1.74 million households were registered forMSP and had valid postal codes for linkage with area-based income strata. This cohort accounts for 95% of thetotal population in the province. Of these households,1.36 million were registered for the Fair PharmaCare pro-gram. Cross-tabulations of the household-level and area-based income measures are shown in Table 1, where NRindicates the percentage of households in each area-baseddecile who were not registered for the Fair PharmaCareprogram at the time of data collection. This table confirmsthat rates of participation with the income-based programare lower in higher income neighbourhoods. This concen-tration of low incomes for the household-level incomevariable is because the registration for income-based drugcoverage involves a degree of self-selection bias. To adjustfor this, our tables below present a "best case" scenariowherein all non-registered households are assigned ahypothetical household-level income variable that isidentical to their area-level income.Table 2 shows the level of discrepancy between the house-hold-level and area-based income measures. The area-based measures classify 15.6% of senior households and14.9% of non-senior households as being within the sameincome-decile as is determined by tax-reported householdincome. Approximately a third of non-senior householdsand two fifths of senior households are classified by area-based measures to be within one decile of the classifica-tion based on household-level incomes. In the "best-case"scenario, just over half of non-seniors and approximately43% of seniors are within one decile of their household-level income.Statistical evidence of the disagreement between incomemeasures can be found in Table 3. The Spearman's corre-lations between the actual household-level income andarea-based measures are always less than 0.40, suggestinglittle agreement. The kappa coefficient of non-random,complete agreement never exceeds 0.31 indicating very lit-tle complete agreement between area-based and actualhousehold-level deciles even under the assumption ofperfect correlation between area-based and household-level measures for all non-registrants. Again, when exam-ining the weighted kappa coefficients, incorporating par-tial agreement, we see that they never exceed 0.5, even inthe best-case scenario.To examine whether these discrepancies result in anymeaningful differences in an applied research context, westart by examining the distribution of total prescriptiondrug expenditures by income deciles stratified by seniorand non-senior households, first using household-levelCRA validated income and then using aggregate neigh-bourhood level income (Table 4). Table 4 indicates thattotal prescription drug expenditures appear more equallyTable 1: Entire BC population, 2003. Agreement between household-level validated income deciles and area-based income decilesHousehold-level validated income decile1 2 3 4 5 6 7 8 9 10 NR totalArea-based income decile 1 20.1 13.0 13.1 10.8 7.7 5.7 5.0 3.6 2.6 1.7 16.6 1002 10.8 10.5 10.7 10.4 9.4 8.2 7.6 5.9 4.6 2.7 19.2 1003 9.1 9.2 9.7 9.7 9.5 8.7 8.1 6.8 5.7 3.5 20.0 1004 7.8 8.2 8.9 9.2 9.2 8.8 8.5 7.7 6.7 4.3 20.8 1005 6.8 7.6 7.7 8.3 8.6 8.7 8.6 8.5 7.8 5.4 22.1 1006 5.7 7.0 7.1 7.5 8.2 8.6 8.7 8.8 8.9 6.8 22.8 1007 4.9 6.5 6.3 6.8 7.7 8.4 8.7 9.3 9.7 8.6 23.2 1008 4.7 6.0 5.6 6.0 6.8 7.9 8.5 9.6 10.8 11.0 23.2 1009 4.0 5.4 5.0 5.2 6.0 7.4 8.1 9.6 11.1 14.1 24.1 10010 4.2 4.8 4.2 4.2 4.9 5.9 6.5 8.3 10.2 20.1 26.6 100Page 3 of 7(page number not for citation purposes)Note: NR indicates individuals who were not registered for Fair PharmaCare at the time of data collection.BMC Health Services Research 2008, 8:79 http://www.biomedcentral.com/1472-6963/8/79distributed when we rank households by neighbourhoodincome than by household-level income, suggesting thatneighbourhood level income masks variation in theunderlying household-level income variable.In Table 5 we estimate the effect of household income ontotal prescription drug expenditures by using both house-hold-level income and neighbourhood level income inseparate regressions. The dummy variable for the highestincome decile was not included in the regression; thus,the coefficients can be interpreted as the difference in totalprescription drug costs between each income decile andthe highest income decile. The regression results alsoreflect the pattern noted in Table 4. While the signs neverdiffer, the household-level variables pick up a substan-tially larger coefficient than the corresponding neighbour-hood-level variable. This again suggests that theneighbourhood-level variables are smoothing the distri-bution of total prescription drug expenditures acrossincome deciles. While the coefficients on income decilesdiffer substantially between the two models, it is interest-ing to note that the coefficients on presence of seniors andhousehold size do not. Both coefficients are in the samedirection and are of the same magnitude indicating thatthe difference in income variable does not have a largeeffect on other coefficients in the model. The model basedon household-level income also reports a higher adjustedR[2] statistic than that using the area-based measure, indi-cating that the goodness of fit is higher in the regressionusing household-level variables. We also find that theinclusion of covariates in the model does not attenuatethe bias between the variables substantially (Table 5).DiscussionWe found a sufficient level of discrepancy between thearea-based and household-level income measures. Usingvalidated household income as the standard, area-basedmeasures misclassified the income decile for eighty-fivepercent or more of the households in the data. We alsofound that these discrepancies did affect the size of coeffi-cients in regression analyses, suggesting that very differentconclusions can be reached regarding the 'same' issuedepending on which income variable we use. Thus, theseresults indicate that, at least in some contexts, the choiceof neighbourhood versus household income can driveconclusions in an important way. Our results are consist-ent with a large amount of work indicating substantial dis-crepancy between area-based and household SESmeasures[2,6,8,10].There are also a couple of important caveats. The first isthat our study did not examine the inclusion of income assimply one of several control variables, but rather onlylooked at the difference between household-level andTable 2: Percentage of discrepancy by decile between area-based and household-level income measuresArea-based Measure Group None One Two Three Four Five Six Seven Eight NineActual household-level income All 14.9 22.5 18.5 14.5 10.8 7.7 5.3 3.3 1.8 0.7Non-seniors 14.9 21.8 18.2 14.7 11.2 7.9 5.4 3.4 1.8 0.7Seniors 15.6 23.9 19.1 14.4 10.4 7.2 4.7 2.7 1.4 0.6Best-case scenario All 33.4 18.3 14.4 11.4 8.3 5.4 4.1 2.6 1.4 0.7Non-seniors 37.8 16.0 13.3 10.7 8.2 5.8 3.9 2.5 1.3 0.5Seniors 20.2 22.7 18.0 13.6 9.8 6.8 4.4 2.6 1.3 0.6Note: In Best-case scenario, non-registered households are assigned a hypothetical household-level income decile that is identical to their area-based measure.Table 3: Spearman's correlation, Kappa and weighted Kappa coefficients for the association between the area-based income measures and the household income measureArea-based measure Group rs Kappa Weighted KappaActual household-level income All 0.322 0.055 0.322Non-seniors 0.316 0.054 0.315Seniors 0.382 0.060 0.382Best-case Scenario (including non-registrants) All 0.469 0.260 0.469Non-seniors 0.497 0.309 0.496Seniors 0.420 0.113 0.420Note: r = Spearman's correlation coefficient.Page 4 of 7(page number not for citation purposes)s In Best-Case Scenario, non-registered households are assigned a hypothetical household-level income decile that is identical to their area-based measure.BMC Health Services Research 2008, 8:79 http://www.biomedcentral.com/1472-6963/8/79area-level income when applied as the primary variable ofinterest. Thus, results cannot be extended to the use ofincome as a control in much larger regression equations.Second, these results are not meant to suggest that the useof neighbourhood income is inferior in all contexts. Anauthor particularly concerned with measuring permanentincome free of yearly fluctuations may find that neigh-bourhood income provides a better measure. When meas-have better access to care than other similar low-incomefamilies simply because of where they live. Thus, an argu-ment could be made for including both measures in thistype of work.While the level of agreement between area-based andhousehold-level SES measures has frequently been stud-ied, our work adds to the knowledge base for several rea-Table 4: Total drug costs by income decileEntire BC population Non-senior population Senior populationMean total drug costsPercent of total drug costsMean total drug costsPercent of total drug costsMean total drug costsPercent of total drug costsDeciles measured by CRA validated income1 ($900–3,750) 743.27 10.39 826.46 15.24 894.61 6.992 ($3750–12,000) 615.16 8.60 456.26 8.41 942.67 7.373 ($12,000–16,000)605.19 8.46 332.54 6.13 1059.92 8.284 ($16,000–20,000)681.31 9.53 334.55 6.17 1178.95 9.215 ($20,000–29,375)534.48 7.47 358.74 6.61 1280.50 10.016 ($29,375–39,584)653.91 9.14 411.01 7.58 1352.22 10.577 ($29,584–51,250)705.15 9.86 503.90 9.29 1420.86 11.118 ($51,250–67,917)821.07 11.48 654.20 12.06 1499.01 11.729 ($67,917–91,667)881.27 12.32 741.34 13.67 1565.45 12.2410 ($91,667–475,000)912.01 12.75 804.41 14.83 1599.72 12.50100 100 100Deciles measured by neighbourhood income1 ($4,200–20,100) 726.63 10.16 615.32 11.35 1109.80 8.672 ($20,100–23,800)653.33 9.13 499.71 9.21 1189.72 9.303 ($23,800–26,500)697.09 9.75 511.38 9.43 1245.68 9.744 ($26,500–29,00) 700.24 9.79 510.90 9.42 1273.95 9.965 ($29,00–31,300) 707.28 9.89 523.97 9.66 1271.18 9.946 $31,300–33,900) 711.68 9.95 531.67 9.80 1309.01 10.237 ($33,900–36,900)720.97 10.08 540.98 9.97 1323.70 10.358 ($36,900–41,300)722.07 10.09 545.90 10.07 1343.11 10.509 ($41,300–48,900)728.69 10.19 554.05 10.22 1349.44 10.5510 ($48,900–310,900)784.83 10.97 589.86 10.87 1378.30 10.77100 100 100Note: All numbers are based on the best-case scenario in which all non-registered households are assigned a hypothetical individual-level income decile that is identical to their area-based measure.Page 5 of 7(page number not for citation purposes)uring access to health care, it might also be true that low-income families living in high-income neighbourhoodssons. It encompasses a larger number of Canadians, asample of 78% of all households in British Columbia, ofBMC Health Services Research 2008, 8:79 http://www.biomedcentral.com/1472-6963/8/79which 95% of all senior households are analyzed. Also,while other studies have tended to compare area-basedmeasures to household-level survey data[6,8,9] or havecompared two or more different sized area-based meas-ures[10,11] we have used highly reliable household-levelincome data validated with the Canada Revenue Agency.Therefore, we have been able to avoid all self-reportingbias, we have a great deal of confidence in our household-level income variable, and we have been able to analyzealmost an entire population of a Canadian province.ConclusionWhile many authors have argued that household-levelincome should be used whenever possible, census-basedaggregate measures will continue to be necessary forhealth research until household-level data become morereadily available. Two suggestions can be made based onthese research results. The first is that researchers shouldbe cautious when interpreting the results of studies usingaggregate measures as proxies for individual and house-hold income. Area-based measures are approximationsthat are best suited to investigating major differences inincomes (e.g., differences of two or more quintiles) or tostudying context in which someone lives rather than theirspecific income. The second suggestion is perhaps obvi-ous to researchers but important for governments and sta-tistical agencies to fully understand: access to reliableindividual-level income/socio-economic data, as well asthe neighbourhood level income data that is currentlyavailable, would unambiguously improve health researchand therefore the evidence on which health and socialpolicy would ideally rest.Competing interestsThe author(s) declare that they have no competing inter-ests.Authors' contributionsGH participated in conception of the study and studydesign, performed the statistical analysis and drafted themanuscript. SM participated in conception of the studyand study design and participated in drafting the manu-script. Both authors read and approved the final manu-script.AcknowledgementsThis research was supported by a CIHR operating grant. Steve Morgan is supported by a New Investigator award from the Canadian Institutes of Health Research (CIHR) and Scholar Award from the MSFHR.References1. Adler NE, Boyce WT, Chesney MA, Folkman S, Syme SL: Socioeco-nomic inequalities in health. No easy solution.  JAMA 1993,269:3140-3145.2. Braveman PA, Cubbin C, Egerter S, Chideya S, Marchi KS, Metzler M,Posner S: Socioeconomic Status in Health Research One SizeDoes Not Fit All.  JAMA 2005, 294:2879-2888.3. Marmot MG, Rose G, Shipley M, Hamilton PJ: Employment gradeand coronary heart disease in British civil servants.  Br Med J1978, 32(4):244-9.4. van Doorslaer E, Masseria C, Koolman X: Inequalities in access tomedical care by income in developed countries.  Can Med AssocJ 2006, 174:177-183.5. Culyer AJ: Health, Health Expenditures and Equity.  Universityof York, Centre for Health Economics; 1991. 6. Geronimus AT, Bound J, Neidert LJ: On the Validity of UsingCensus Geocode Characteristics to Proxy Individual Socioe-conomic Characteristics.  J Am Stat Assoc 1996, 91:529-537.7. Last JM: A Dictionary of Epidemiology.  Oxford: Oxford Univer-sity Press; 1995. 8. Demissie K, Hanley JA, Menzies D, Joseph L, Ernst P: Agreement inmeasuring socio-economic status: area-based versus individ-ual measures.  Chronic Dis Can 2000, 21(1):1-7.9. Diez-Roux AV, Kiefe CI, Jacobs DR Jr, Haan M, Jackson SA, Nieto FJ,Paton CC, Schulz R: Area characteristics and individual-levelsocioeconomic position indicators in three population-basedepidemiologic studies.  Ann Epidemiol 2001, 11:395-405.10. Geronimus AT, Bound J: Use of census-based aggregate varia-bles to proxy for socioeconomic group: evidence fromnational samples.  Am J Epidemiol 1998, 148:475-486.11. Southern DA, Ghali WA, Faris PD, Norris CM, Galbraith PD, GrahamMM, Knudtson ML: Misclassification of income quintilesTable 5: Results for the regression of dummy variables indicating income decile against total drug costsExplanatory Variables Household income Neighborhood Income Household income (without covariates)Neighborhood Income (without covariates)Income decile 1 -169.00 (-30.27) -59.162 (-10.57) -68.22 (-9.87) -14.57 (-2.65)Income decile 2 -297.08 (-53.20) -128.26 (-22.92) -405.50 (-58.67) -96.59 (-17.61)Income decile 3 -305.05 (-54.63) -91.67 (-16.38) -289.43 (-41.87) -80.34 (-14.65)Income decile 4 -232.39 (-41.62) -85.20 (-15.22) -179.73 (-26.00) -72.91 (-13.29)Income decile 5 -378.51 (-67.78) -77.08 (-13.77) -41.74 (-6.04) -67.03 (-12.22)Income decile 6 -257.56 (-46.13) -74.33 (-13.28) -40.48 (-5.85) -54.49 (-9.94)Income decile 7 -207.80 (-37.21) -64.83 (-11.58) -34.68 (-5.02) -44.99 (-8.21)Income decile 8 -89.99 (-16.12) -62.60 (-11.18) -53.00 (7.67) -39.09 (-7.13)Income decile 9 -32.60 (-5.84) -57.38 (-10.25) -14.63 (-2.12) -34.05 (-6.21)Presence of seniors 787.81 (250.77) 790.37 (267.87) Not included Not includedHousehold size 105.43 (89.26) 104.65 (92.14) Not included Not includedAdjusted R2 0.101 0.06 0.01 0.01Note: T statistics are in parentheses. All coefficients are significant at 95% confidence interval.Page 6 of 7(page number not for citation purposes)derived from area-based measures: A comparison of enu-meration area and forward sortation area.  Can J Public Health2002, 93:465-469.Publish with BioMed Central   and  every scientist can read your work free of charge"BioMed Central will be the most significant development for disseminating the results of biomedical research in our lifetime."Sir Paul Nurse, Cancer Research UKYour research papers will be:available free of charge to the entire biomedical communitypeer reviewed and published immediately upon acceptancecited in PubMed and archived on PubMed Central BMC Health Services Research 2008, 8:79 http://www.biomedcentral.com/1472-6963/8/79Pre-publication historyThe pre-publication history for this paper can be accessedhere:http://www.biomedcentral.com/1472-6963/8/79/prepubyours — you keep the copyrightSubmit your manuscript here:http://www.biomedcentral.com/info/publishing_adv.aspBioMedcentralPage 7 of 7(page number not for citation purposes)


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items