Reserch Articles Are-Level Socioeconomic Disdvntge nd Severe Pulmonry Tuberculosis: U.S., 2000 2008 Eyl Oren, PhD,b Mshiro Nrit, MD b,c Chrles ln, MD Jonthn Myer, PhD ABSTRACT Objectives. Lower socioeconomic sttus (SES) is ssocited with incresed risk of tuberculosis (TB) nd dignostic delys, but the extent to which this ssocition reflects n underlying grdient in dvnced sttus of pulmonry TB is unknown. We conducted multilevel retrospective cohort nlysis exmining the reltionship between socioeconomic chrcteristics nd pulmonry TB disese sttus, s mesured vi sputum smers nd chest rdiogrphy results. Methods. We included 862 incident TB ptients reported in King County, Wshington, from 2000 2008. We bstrcted ptient-level mesures from chrts nd surveillnce dt. We obtined socioeconomic chrcteristics of TB ptients, s well s those of the res where TB ptients lived, from the 2000 U.S. Census. A socioeconomic position (SEP) index ws derived to mesure SES. Results. Of those with known results, 814 of 849 ptients (96%) displyed bnorml rdiogrphy findings. A totl of 239 grded ptients (39%) hd positive smers, 136 (57%) of whom hd grdes of moderte (31) or numerous (41) cid-fst bcilli. In undjusted nlyses, ptients living in lower SEP res did not pper to hve higher probbilities of more dvnced disese. In multivrite models djusting for individul demogrphic nd socioeconomic mesures, s well s re-bsed demogrphic vribles, block-group SEP ws not significntly ssocited with more dvnced pulmonry disese. Conclusions. Lower SEP ws not significntly ssocited with greter pulmonry disese severity fter controlling for individul ge, rce, sex, nd origin, nd block-group rce, ethnicity, nd origin. These findings suggest tht the severity of pulmonry TB t dignosis is not synonymous with delyed dignosis. University of Wshington, Deprtment of Epidemiology, Settle, WA b Public Helth Settle & King County, Tuberculosis Control Progrm, Settle, WA c University of Wshington, Division of Pulmonry nd Criticl Cre Medicine, Settle, WA Address correspondence to: Eyl Oren, PhD, Public Helth Settle & King County, Tuberculosis Control Progrm, HMC-PH-0100, 325 Ninth Ave., Settle, WA 98104; tel. 206-744-2196; e-mil <eoren@uw.edu>. 2013 Assocition of Schools of Public Helth 99
100 Reserch Articles Although tuberculosis (TB) incidence continues to decline in the United Sttes, the proportion of dvnced pulmonry TB, defined s smer-positive or cvitry disese, hs been incresing. 1,2 Advnced clinicl presenttion my result from delyed dignosis nd tretment nd my led to greter infectivity nd likelihood of trnsmission within community. 3,4 Lower socioeconomic sttus (SES) hs been linked to more severe disese sttus for vriety of diseses including cystic fibrosis, 5 srcoidosis, 6 subclinicl coronry hert disese, 7 cncers, 8,9 nd pulmonry fibrosis. 10 While the presence of other comorbidities, poor ccess to cre, substnce buse, low income, eduction level, nd lck of insurnce 11 16 constitute risk fctors for delys in TB dignosis, little work hs been done to chrcterize the ssocition between SES nd dvnced pulmonry TB disese, s more dvnced disese is often seen s chrcteristic of dignostic dely. Furthermore, much of the work exmining dignostic dely hs been exmined outside the U.S. Ares in which people live re likely to hve differentil ccess to cre, including proximity, cost, nd presence of public clinics nd trnsporttion. 17,18 Disese sttus is likely to be impcted by such vritions in re-level fctors nd, in prticulr, by vritions in re-level SES cross neighborhoods. Using TB cse registry dt on incident TB ptients combined with chrt reviews, we explored the reltionship between individul ptient demogrphic nd SES ttributes, in combintion with re-level socil chrcteristics, nd two TB severity outcomes t dignosis lung cvittion nd smer-positive cidfst bcilli (AFB) in sputum smers. These mesures hve been linked to impired pulmonry function, TB tretment filure, or deth 19,20 nd represent lter disese stges. 21,22 Specificlly, this study ws designed to ssess whether severity of pulmonry TB disese ws positively ssocited with re-bsed socioeconomic disdvntge within King County, Wshington. By exmining socioeconomic nd demogrphic chrcteristics of block groups, our hope ws to identify those fctors tht might ply n importnt role in predicting disese severity t dignosis. Such findings could provide insight into pthwys by which re-level SES independently ffects pulmonry disese severity. METHODS Study popultion nd setting A totl of 862 incident pulmonry TB ptients were reported in King County from Jnury 1, 2000, to December 31, 2008. Criteri for dignosis of pulmonry TB ptients met either Centers for Disese Control nd Prevention (CDC) lbortory criteri for sputum culture positive dignosis or the clinicl cse definition, which includes either n bnorml chest rdiogrph or other signs nd symptoms comptible with TB. 23 Individuls hd either exclusively pulmonry disese or pulmonry involvement. Due to considertions regrding ge of reserch consent, ll models excluded minors ( 18 yers of ge). Included ptients represented 380 block groups within King County. A census block group ws defined s cluster of census blocks hving the sme first digit of their four-digit identifying numbers within census trct. 24 Block groups hve previously been vlidted s n informtive sptil scle t which to report socioeconomic dt. 25 Study design The nlysis used retrospective cohort design, merging reporting nd chrt dt for TB ptients nd U.S. block-group-level census dt for residents of King County. Dt sources Mesurement of socioeconomic position. A socioeconomic position (SEP) index ws constructed combining dt on six singulr SES mesures: percentge of the popultion who were working clss, were unemployed, were living in poverty, hd less thn high school eduction, nd owned expensive homes, s well s medin household income. To construct the score, we gve ech vrible stndrdized z-score, which ws the sum of ll block-group vlues with SEP dt (n51,576), minus the men sum, divided by the stndrd devition, nd then summed the individul z-scores. Vribles included for ech block group were tken from the 2000 U.S. Census Summry Tpe File-3 nd were consistent with previously vlidted composite mesure used in the Public Helth Disprities Geocoding Project. The Geocoding Project developed this mesure bsed on fctor nlysis of 11 individul SES fctors using rnk vlues of the Census dt. 25 The SEP index ws modeled s four-level ctegoricl vrible, using qurtiles in the block-group distribution s cutoffs, with the highest qurtiles representing the welthiest block groups. Mesurement of disese severity. We chose two vilble chrcteristics of dvnced pulmonry TB t dignosis from 2000 2008: grde of sputum smer nd chest rdiogrphic bnormlities. Sputum smers were quntified using fluorochrome stins nd ctegorized using sputum smer grdes in n ordinl fshion s negtive (no AFB seen), 11 (rre), 21 (few), 31 (moderte),
Socioeconomic Disdvntge nd TB: 2000 2008 101 nd 41 (numerous), depending on AFB lod. Smers were lso dichotomized s positive or negtive. Stndrd posterior-nterior chest rdiogrphs were ctegorized ordinlly s norml, bnorml non-cvitry, or cvitry in the ptient medicl chrt by the Public Helth Settle & King County TB Control Progrm medicl director, with ptients with norml rdiogrphs serving s the referent ctegory. Rdiogrphs were lso seprtely ctegorized s to whether there ws unilterl or bilterl pulmonry involvement. Both rdiogrphy nd smer results were obtined from the Tuberculosis Informtion Mngement System (TIMS) 26 nd supplemented with vilble dt from ptient medicl chrts with quntittive smer grde. Smer grde ws obtined for the 76% of ptients with complete medicl records. Anthropometric nd psychosocil individul mesures. The following individul mesures from the Ntionl Tuberculosis Surveillnce System 23 were used: rce, sex, ge, ethnicity, foreign birth, homelessness, humn immunodeficiency virus (HIV) sttus, nd provider type. In ddition, n experienced nurse mesured height nd weight to compute body mss index (BMI). Prticipnts reported on stndrdized clinic forms whether or not they were in pid employment t dignosis, s well s their occuption, insurnce, smoking sttus, nd lcohol intke history. Ptient-level vribles were subsequently ggregted by block group. Other re-level mesures. We derived re-bsed covrites from the U.S. Popultion Census 2000, SF1 nd SF3. 27,28 We modeled the proportion of ech block group tht ws blck, Asin, Hispnic, nd foreignborn using qurtiles of ech popultion in ech block group, with the qurtile directly below serving s the referent group. Sttisticl nlysis nd modeling We excluded observtions from univrite nlyses when the percentge unknown or missing ws 2%. We exmined undjusted proportions of individuls in ech SEP index qurtile nd strtified them by both quntittive smer grde nd rdiogrphy results, with percentges dding up to 100 cross SEP index qurtiles. We lso exmined the reltionship between TB sputum smer grde nd rdiology results using Person s Chi-squre test. To exmine re-level influences in ddition to individul ttributes s they relte to vrition in severity of disese, we used multilevel logistic regression models. 29 These clsses of models llowed for nlysis of the ordered outcomes nd ccommodted the hierrchicl dt structure. After building multilevel models of significnt fixed effects, other prmeters were dded, llowing for bseline vrition in disese severity cross block groups. For ordinl models, the forms used were similr, but used n ordered logit model llowing for three responses for rdiogrphy outcomes (norml, bnorml non-cvitry, nd bnorml cvitry) nd five responses for smer grde (negtive, 11, 21, 31, nd 41). For ech outcome, we tested four nested models. Model 1 tested re-bsed socioeconomic qurtiles nd the ssocition with dependent vribles. Model 2 included individul demogrphic fctors (mencentered ge modeled continuously nd s qudrtic term, rce modeled using dummy vribles, sex modeled s binry term, nd foreign birth s binry term) s covrites. Individul-level SES fctors were dditionlly included (homelessness s binry term nd provider type s dummy vrible) in Model 3. Are-level fctors (ethnicity, foreign birth, nd rce) were dded in Model 4 to ssess the contextul effects of Asin nd blck rce, Hispnic ethnicity, nd foreign birth while controlling for individul confounders nd re-level SEP. We used complete cse nlysis such tht the number of ptients with missing covrites excluded from ech model ws the sme. We performed multilevel model building nd estimtion using the Generlized Liner Ltent nd Mixed Models extension of Stt version 10.0. 30 RESULTS Description of TB ptients Tble 1 portrys overll nd nlysis-specific ptient popultion chrcteristics. A totl of 862 TB ptients were included in the initil nlysis, with 65% being mle nd medin ge t dignosis of 44 yers. TB ptients were primrily Asin (40%), blck (24%), or white (26%). More thn 70% of the ptient popultion ws foreign-born. TB risk fctors included homelessness (20%), unemployment prior to dignosis (36%), HIV infection (9% of known results), nd smoking (28% of known results). Ptients were included in subsequent nlyses if sputum specimen (n5806) nd/or chest rdiogrph result (n5849) ws vilble. A totl of 616 ptients were included in subgroup for quntittive smer-grde nlysis, excluding ptients for whom smer grdes were unknown (n5246). The flow chrt shows the ptient inclusion process for ech severity mesure (Figure 1).
102 Reserch Articles Tble 1. Chrcteristics of TB ptients reported in King County, Wshington: 2000 2008 Smer sttus Chest rdiogrphy degree Chrcteristic Popultion Positive Negtive rml Abnorml Cvitry Totl (N) 862 416 390 35 579 235 Sex: mle 556 (64.5) 289 (69.5) 237 (60.8) 19 (54.3) 363 (62.7) 168 (71.5) Age (in yers): men (SD) Age ctegories (in yers) 0 4 5 14 15 24 25 44 45 64 65 44.2 (20.5) 18 (2.1) 14 (1.6) 132 (15.3) 297 (34.5) 236 (27.4) 161 (18.7) 43.4 (19.0) 0 (0.0) 3 (0.7) 76 (18.3) 153 (36.8) 119 (28.6) 65 (15.6) 45.5 (20.3) 3 (0.8) 8 (2.1) 55 (14.1) 137 (35.1) 106 (27.2) 81 (20.8) 39.7 (14.9) 1 (2.9) 0 (0.0) 4 (11.4) 20 (57.1) 9 (25.7) 1 (2.9) 45.8 (21.6) 16 (2.8) 12 (2.1) 78 (13.5) 183 (31.6) 157 (27.1) 133 (23.0) 40.7 (17.8) 1 (0.4) 2 (0.9) 49 (20.9) 90 (38.3) 68 (28.9) 25 (10.6) Rce Americn Indin Asin Blck Ntive Hwiin White Multiple rces 49 (5.7) 353 (40.1) 205 (23.8) 19 (2.2) 224 (26.0) 2 (0.2) 31 (7.5) 151 (36.3) 104 (25.0) 9 (2.2) 115 (27.6) 1 (0.2) 18 (4.6) 175 (44.9) 94 (24.1) 9 (2.3) 92 (23.6) 1 (0.3) 4 (11.4) 8 (22.9) 12 (34.3) 1 (2.9) 20 (28.6) 0 (0.0) 36 (6.2) 247 (42.7) 130 (22.5) 13 (2.3) 148 (25.6) 2 (0.4) 8 (3.4) 95 (40.4) 59 (25.1) 5 (2.1) 65 (27.7) 0 (0.0) Ethnicity Hispnic b 96 (11.1) 60 (14.4) 31 (8.0) 6 (17.1) 53 (9.2) 36 (15.3) Country of origin U.S.-born Foreign-born c Time from U.S. rrivl to TB dignosis (in yers) d 0 4 5 9 10 19 20 HIV sttus Negtive Positive Unknown/missing e Homeless within pst yer Insurnce Unemployed within pst 24 months Injecting drug use within pst yer n-injecting drug use within pst yer Excess lcohol use within pst yer f 250 (29.0) 610 (70.8) 235 (38.5) 91 (14.9) 135 (22.1) 101 (16.6) 48 (7.9) 634 (73.9) 62 (7.2) 162 (18.9) 688 (79.8) 169 (19.6) 277 (32.1) 144 (16.7) 441 (51.2) 314 (36.4) 548 (63.6) 806 (93.5) 26 (3.0) 30 (3.5) 757 (93.5) 70 (8.1) 35 (4.1) 704 (81.7) 131 (15.2) 27 (3.2) 118 (28.4) 297 (71.4) 110 (37.0) 54 (18.2) 62 (20.9) 53 (17.9) 18 (3.1) 319 (76.7) 34 (8.2) 62 (15.1) 310 (74.5) 106 (25.5) 144 (34.4) 77 (18.4) 198 (47.3) 145 (34.9) 271 (65.1) 384 (92.3) 17 (4.1) 15 (3.6) 353 (84.9) 45 (10.8) 18 (4.3) 314 (75.5) 87 (20.9) 15 (3.6) 102 (26.2) 287 (73.6) 121 (42.2) 36 (12.5) 65 (22.7) 42 (14.6) 23 (8.0) 293 (75.1) 26 (6.7) 71 (18.2) 327 (83.9) 62 (15.9) 198 (33.5) 105 (17.7) 289 (48.8) 144 (36.9) 246 (63.1) 374 (95.9) 8 (2.1) 8 (1.0) 357 (91.5) 23 (5.9) 10 (2.6) 343 (88.0) 43 (11.0) 4 (1.0) 15 (42.9) 20 (57.1) 8 (40.0) 2 (10.0) 3 (15.0) 5 (25.0) 2 (10.0) 23 (65.7) 9 (25.7) 3 (8.6) 19 (54.3) 15 (45.7) 13 (37.1) 2 (5.7) 20 (57.1) 15 (42.9) 20 (57.1) 30 (85.7) 3 (8.6) 2 (5.7) 26 (74.3) 7 (20.0) 2 (5.7) 25 (71.4) 7 (20.0) 3 (8.6) 165 (28.5) 413 (71.3) 165 (39.7) 54 (13.1) 96 (23.2) 66 (16.0) 33 (8.0) 405 (70.0) 44 (7.6) 130 (22.4) 479 (82.7) 99 (17.2) 195 (33.7) 90 (15.5) 294 (50.8) 227 (39.2) 352 (60.8) 547 (94.5) 17 (2.9) 2 (2.6) 524 (90.5) 38 (6.6) 17 (2.9) 490 (84.6) 79 (13.6) 10 (1.7) 63 (26.8) 171 (72.8) 60 (35.1) 35 (20.5) 35 (20.5) 30 (17.5) 11 (6.4) 201 (85.5) 7 (3.0) 27 (11.5) 182 (77.5) 53 (22.6) 65 (27.7) 51 (21.7) 119 (50.6) 70 (29.8) 165 (70.2) 220 (93.6) 6 (2.6) 9 (3.8) 199 (84.7) 24 (10.2) 12 (5.1) 181 (77.0) 45 (19.2) 9 (3.8) continued on p. 103
Socioeconomic Disdvntge nd TB: 2000 2008 103 Tble 1 (continued). Chrcteristics of TB ptients reported in King County, Wshington: 2000 2008 Smer sttus Chest rdiogrphy degree Chrcteristic Popultion Positive Negtive rml Abnorml Cvitry Smoking history Chest rdiogrphic result rml Abnorml, non-cvitry Abnorml, cvitry t done Bilterl lung involvement Unknown Sputum smer result g Positive Negtive t done Provider type Helth deprtment Privte provider Both 348 (40.4) 135 (15.7) 379 (44.0) 35 (4.1) 579 (66.0) 235 (27.3) 3 (0.4) 288 (33.4) 183 (21.2) 391 (45.4) 416 (48.2) 390 (45.2) 48 (5.6) 686 (79.6) 65 (7.5) 101 (11.7) 177 (42.2) 75 (17.9) 167 (39.9) 7 (1.7) 211 (50.7) 193 (46.4) 1 (0.2) 134 (32.0) 114 (27.2) 171 (40.8) 338 (81.3) 26 (6.3) 49 (11.8) 273 (46.1) 74 (12.5) 245 (41.4) 22 (5.6) 330 (84.6) 36 (9.2) 1 (0.3) 254 (42.9) 85 (14.4) 253 (42.7) 313 (80.3) 29 (7.4) 45 (11.5) 12 (34.3) 5 (14.3) 18 (51.4) 16 (45.7) 1 (2.9) 18 (51.4) 7 (20.0) 22 (62.9) 4 (11.4) 27 (77.1) 2 (5.7) 6 (17.1) 245 (42.3) 84 (14.5) 250 (43.2) 209 (36.1) 108 (18.7) 262 (45.3) 211 (36.4) 330 (57.0) 37 (6.4) 454 (78.8) 52 (4.7) 70 (9.2) 86 (36.6) 46 (19.6) 103 (43.8) 60 (25.5) 72 (30.6) 103 (43.8) 193 (82.1) 36 (15.3) 5 (2.1) 198 (80.6) 11 (7.7) 23 (11.7) Percentges my not totl 100 due to rounding nd exclusion of unknown when 2%. b People of Hispnic ethnicity my be of ny rce or multiple rces. c Foreign-born includes people born outside the U.S., Americn Smo, the Federted Sttes of Micronesi, Gum, the Republic of the Mrshll Islnds, Midwy Islnd, the Commonwelth of the rthern Mrin Islnds, Puerto Rico, the Republic of Plu, the U.S. Virgin Islnds, nd U.S. minor nd outlying Pcific islnds. d Among foreign-born ptients e Unknown or missing includes indeterminte, refused, not offered, test done but unknown, nd unknown. f Excess lcohol use is $5 drinks on sme occsion on ech of 5 dys in pst 30 dys. g From the Ntionl Tuberculosis Surveillnce System; smers not done/unknown ccount for 56 (6%) of totl smer results. TB 5 tuberculosis SD 5 stndrd devition HIV 5 humn immunodeficiency virus 5 not vilble Block-group chrcteristics The 380 block groups included in the nlysis were more likely to contin individuls reporting blck or Asin descent s well s Hispnic ethnicity compred with medin vlues in King County. Additionlly, the medin proportion of foreign-born individuls in these block groups ws more thn 1.5 times s high s the King County medin (Tble 2). Disese severity findings Of those with known results, 96% (814/849) of ptients displyed bnorml rdiogrphy findings, with pproximtely one-third (33%) exhibiting extensive bilterl lung involvement. For the quntittive smer-grde outcome nlysis, 39% (239/616) of grded ptients hd positive smers, of whom 57% hd grdes of moderte (31) or numerous (41) AFB. Higher grdes of smer nd cvitry rdiogrphs were positively correlted, where greter proportions of cvitry x-rys were observed with progressively higher smer grde (p 0.001). Bilterl lung disese ws significntly ssocited with both higher sputum smer grde (p 0.001) nd cvitry disese (p 0.001) (dt not shown). Eighty-one percent of TB ptients resided in block groups in the lowest two SEP index qurtiles. In undjusted nlyses, ptients living in res with higher levels of deprivtion did not hve sttisticlly higher probbilities of severe rdiogrphs or higher smer grde (Figures 2 nd 3).
104 Reserch Articles Figure 1. Flow digrm detiling inclusion criteri in n nlysis of socioeconomic disdvntge nd pulmonry tuberculosis severity: King County, Wshington, 2000 2008 SEP 5 socioeconomic position Tble 2. Chrcteristics of 380 block groups included in n nlysis of socioeconomic disdvntge nd pulmonry tuberculosis severity in King County, Wshington, bsed on 2000 U.S. Census dt Vrible Medin Men SD Rnge King County medin Demogrphics Popultion size (N) 1,130 1,211 545 246 4,721 1,011 Rce n-hispnic white (percent) 69.1 63.3 22.7 3.1 97.5 79.7 n-hispnic Asin (percent) 11.3 14.9 14.0 0.0 73.2 7.1 n-hispnic blck (percent) 4.7 8.5 10.6 0.0 56.3 1.8 Hispnic ethnicity (percent) 4.9 7.2 7.1 0.0 44.4 3.5 Foreign-born (percent) b 17.4 20.2 12.8 0.0 62.4 11.7 Socioeconomics High school eduction (percent) 8.7 12.2 10.7 0.0 59.7 6.9 Unemployment (percent) 2.9 3.5 3.1 0.0 26.4 2.6 Medin household income (in U.S. dollrs) $50,357 $52,741 $21,545 $7,382 $146,129 $56,691 Poverty (percent) 7.5 10.5 10.2 0.0 62.7 5.5 Working clss (percent) 56.1 55.5 13.8 14.0 85.4 51.1 Home ownership (percent) 65.1 61.2 27.0 0.0 100.0 73.5 Tuberculosis mesures Men ptients per block group (per yer) 0.1 0.2 0.3 0.1 3.9 0.0 Incidence rte per block group (per 100,000 person-yers) 14.4 22.1 29.8 3.5 334.4 0.0 King County medin reflects ll block groups with socioeconomic sttus vribles vilble (n51,576). b Excluding U.S. territories nd those born brod to U.S. prents SD 5 stndrd devition
Socioeconomic Disdvntge nd TB: 2000 2008 105 Figure 2. Proportion of tuberculosis ptients in qurtiles of block-group SES, by sputum smer grde: King County, Wshington, 2000 2008 Excludes smer grdes tht were not done or unknown, s well s missing SES SES 5 socioeconomic sttus Multivrite nlyses Chest rdiogrph model. In undjusted nlyses, the bseline model indicted tht individuls living in lower SEP index neighborhoods did not hve more severe x-ry presenttion, with the odds rtio (OR) of more severe disese not significntly incresed in the lowest s compred with the highest qurtile (OR50.95, 95% confidence intervl [CI] 0.53, 1.70; p50.935) (Tble 3, Model 1). In ddition, in individul-level multivrite models (Tble 3, Models 2 nd 3), inclusion of demogrphic nd SES covrites did not significntly lter the ssocition between SEP index qurtile nd disese severity, lthough foreign birth decresed nd white rce incresed the odds for more severe presenttion. Inclusion of individul insurnce sttus nd behviorl risk fctors (i.e., drug nd lcohol use, smoking, nd BMI) on smller vilble smples did not chnge the observed ssocition (dt not shown). In multivrite nlyses restricted to non-hiv-infected individuls, no significnt chnges were observed in SEP effect estimtes on severity. Comprison of coefficients from Models 3 nd 4 did not show substntil chnge in the SEP-TB ssocition when other re-level influences were dded (Tble 3, Model 4). There ws modest chnge in the strength of the effect, but direction nd mgnitude of the ssocitions were consistent cross the two models. Of four re-level vribles exmined t the block-group level in ddition to SEP index, none remined sttisticlly significnt in the multilevel model (Tble 3, Model 4). Are-level vribles explined little between-block group vrince, such tht only 9% of vrince in severity ws ttributble to the block group. Sputum smer models. As with the rdiogrph findings, lower SES qurtiles were not ssocited with higher smer grde with ny of the models run. In models exmining binry positive/negtive smer outcomes, positive smer ws not significntly ssocited with SEP qurtile, nd this lck of reltionship remined fter controlling for demogrphic, individul SES, nd rebsed demogrphic fctors. Homelessness ws linked to higher odds of positive sputum smers but did not chnge the observed SEP-smer estimtes. The reltive contribution of ech of the individul-level min effects ws similr in both sputum smer models, suggesting tht re-level fctors did not diminish the effect of individul-level influences. ne of the between-block group vrince in the probbility of hving positive smer result ws ccounted for by block-group SEP. However, when individul-level ge, rce, sex, nd origin were dded in Model 2, the vrince incresed fivefold, indicting tht heterogeneity in disese severity cross block groups ws prtilly ccounted for by underlying demogrphic chrcteristics.
106 Reserch Articles DISCUSSION The results of this reserch, indicting tht residing in res with high levels of poverty is not significntly ssocited with more severe pulmonry disese t dignosis, re noteworthy nd not consistent with previous studies exmining other diseses. 9,31 These findings remined fter ttempting to control for importnt individul-level risk fctors nd re-level mesures nd were consistent cross three mesures of severity. Previous studies of SES nd delyed TB dignoses found tht low income nd poverty constitute risk fctors for delys. 12 14,32 Low eduction level hs previously been described s risk fctor for delys, 15 s hve lrge fmily size, 33 unemployment, 34 nd lck of helth insurnce. 18 However, no studies hve previously documented the direct ssocition between re-bsed SES nd TB disese severity in multilevel model, perhps becuse severity is often seen s representtive of dignostic delys, with longer time to dignosis thought to led to incresed infectiousness s ptients progress to higher bcillry lod on sputum smers nd cvitry disese. 3,35 Previous reports hve hypothesized tht individuls living in poorer SES res present with lter-stge disese becuse of decresed ccess to medicl cre nd screening wreness. 10,36 Yet, possible explntory pthwys for these effects re complex. A previous study found tht lck of employment nd knowledge bout where to obtin cre were closely ssocited with cliniclly significnt dely, rising concerns bout the equity of ccess to cre mong TB ptients. 34 More equitble ccess to cre my occur when the need for cre or severity of illness predicts utiliztion better thn potentil ccess brriers (e.g., ppointment witing time). 37 If ccess to cre were distributed evenly, we would expect tht TB ptients with more severe illness would be more likely to promptly seek medicl cre. But others note tht perceived ccess brriers ppered to explin more of the dely thn did illness severity, suggesting tht subgroups of the TB popultion were fcing inequitble brriers to cre. 34 As such, we might expect tht more severe disese would be impcted by re-level ccess fctors influencing delys. However, while where you live my ply n importnt role in disese incidence nd trnsmission, it my be less importnt fctor in defining the risk of presenting with more severe disese. This finding my be becuse individul SES fctors re often thought to be more closely linked to ccess or usge of helth cre, including longer wit times nd fewer referrls. 38 40 Effects of individul SES my be more pronounced, such tht more comprehensive helth insurnce, greter helth knowledge, nd motivtion Figure 3. Proportion of tuberculosis ptients in qurtiles of block-group SES, by chest rdiogrphy result: King County, Wshington, 2000 2008 Excludes chest rdiogrphs tht were not done or unknown, s well s missing SES SES 5 socioeconomic sttus CXR 5 chest rdiogrphy
Socioeconomic Disdvntge nd TB: 2000 2008 107 Tble 3. Reltive odds of being dignosed with more dvnced tuberculosis ccording to individul nd re-level chrcteristics: King County, Wshington, 2000 2008 Vrible Model 1 OR (95% CI) Model 2 AOR (95% CI) Model 3 AOR b (95% CI) Model 4 AOR c (95% CI) Chest rdiogrphy Highest SEP Medium-high SEP Medium-low SEP Lowest SEP P-vlue Sputum smer grde Highest SEP Medium-high SEP Medium-low SEP Lowest SEP P-vlue Binry sputum smer Highest SEP Medium-high SEP Medium-low SEP Lowest SEP P-vlue 0.80 (0.40, 1.59) 1.21 (0.63, 2.30) 0.95 (0.53, 1.70) 0.935 1.19 (0.69, 2.06) 1.09 (0.62, 1.89) 1.06 (0.65, 1.73) 0.986 1.45 (0.80, 2.64) 1.44 (0.82, 2.50) 1.54 (0.92, 2.63) 0.246 0.83 (0.41, 1.67) 1.18 (0.62, 2.25) 0.88 (0.49, 1.58) 0.680 1.18 (0.68, 2.06) 1.03 (0.59, 1.81) 0.99 (0.60, 1.64) 0.733 1.42 (0.78, 2.59) 1.40 (0.80, 2.45) 1.47 (0.85, 2.53) 0.413 0.85 (0.42, 1.71) 1.22 (0.64, 2.34) 0.91 (0.50, 1.65) 0.764 1.19 (0.68, 2.07) 1.04 (0.59, 1.82) 0.95 (0.57, 1.59) 0.611 1.49 (0.82, 2.70) 1.41 (0.81, 2.46) 1.38 (0.81, 2.37) 0.474 0.84 (0.40, 1.73) 1.18 (0.59, 2.36) 0.88 (0.43, 1.80) 0.863 1.16 (0.66, 2.06) 1.01 (0.54, 1.89) 0.93 (0.49, 1.79) 0.672 1.48 (0.80, 2.73) 1.33 (0.73, 2.43) 1.25 (0.65, 2.43) 0.880 Model 2 djusted for ge, sex, rce, nd foreign birth. b Model 3 djusted for ge, sex, rce, foreign birth, homelessness, nd provider type. c Model 4 djusted for ge, sex, rce, foreign birth, homelessness, provider type, re-level rce, nd ethnicity. OR 5 odds rtio AOR 5 djusted odds rtio SEP 5 socioeconomic position 5 reference group to seek cre ply importnt roles in predicting severity. Indeed, greter proportions of uninsured nd unemployed people were observed in lower SEP qurtile block groups in our study, nd these vribles were significntly ssocited with more severe rdiogrphy results. These dt re consistent with observed correltions between unmet medicl need nd lower income nd lck of insurnce in King County. 40 In undjusted nlyses, the ssocition found between more severe disese nd vrious SES surrogtes hs precedence in the literture. Substnce busers re more likely to hve sputum smer-positive TB disese nd cvitry disese. 41,42 Homelessness is ssocited with smer-positive TB disese nd cvitry disese, 43 nd smoking is ssocited with cvitry lesions. 44 After two months, sputum smer microscopic exmintion is more often positive in dibetic ptients, but we did not exmine dibetes comorbidity in our study popultion. 45 The effect of HIV on TB severity is of prticulr concern. HIV infection my lter the rdiogrphic ppernce of pulmonry TB due to ltered immunity. 46 HIV infection lso promotes rpid progression to ctive TB disese, 47 though its effect on infectiousness remins disputed. 48 Indeed, our results demonstrted tht HIV-infected individuls were more likely to hve bnorml non-cvitry disese. However, in multivrite nlyses restricted to non-infected individuls, no significnt chnges were observed in SES effect estimtes on severity, likely due to smll numbers of HIV-positive people in the nlysis. A recent publiction found tht increses in proportions of dvnced (smer-positive or cvitry disese) pulmonry TB were gretest mong groups with lower rtes of TB, including white, U.S.-born, employed, HIV non-infected, nd non-homeless people. 1 It ws hypothesized tht greter increses in the proportion of dvnced disese mong lower-risk groups were due to lower index of suspicion for TB disese mong ptients nd providers, leding to delys in ccessing tretment nd dignosing disese. Our study results demonstrte the importnce of exmining not only these individul risk fctors, but lso re-level risk fctors for disese. Strengths nd limittions This study hd severl strengths, including the creful ssessment of block-group boundries, vlidtion of the
108 Reserch Articles geocode with the county, incorportion of multilevel models, nd inclusion of re-bsed socioeconomic mesures to exmine SES t both the individul nd re level. Becuse no reltionship ws observed when either smer grde or presence or bsence of positive smer result ws nlyzed, lower block-group SEP did not seem more importnt in distinguishing bcteril lod in the lungs ny more thn it did presence or bsence. This study ws lso subject to severl limittions. One limittion ws the scope of re- nd individul-level vribles studied. There re likely mny re-bsed vribles tht could hve potentilly confounded observed ssocitions between re-bsed SES nd disese severity, s well s relevnt mesures of individul SES tht were unvilble to us. The ltter precluded our bility to ssess reltive impct of re nd individul SES in the prediction of TB severity. While the geogrphic vilbility nd ccessibility of helth-cre services, which my result in differentil dignostic dely, were not included, given King County s predominntly urbn composition, geogrphy ws less likely to hve been strong confounding fctor. Additionlly, these results my not be generlizble to other regions, given tht we djusted only for certin relevnt ptient- nd re-level demogrphic fctors. And becuse residence ws only mesured t TB dignosis, we lso do not know whether residence t previous times could hve been relevnt to the development of disese. Furthermore, unmesured vritions mong block-group risk fctor norms (e.g., verge lcohol intke) could be residully responsible for community contextul effects, but becuse controlling for individul-level risk fctors did not ttenute the block-group SEP effect on disese severity, it seems unlikely tht these fctors would hve n impct on the neighborhood-level SES-severity ssocition. CONCLUSIONS In this study, re-level socil resources were not ssocited with pulmonry TB disese severity t dignosis. These findings re importnt becuse they suggest tht fctors other thn re-level SES my predict severity. At-risk groups should be trgeted for TB interventions regrdless of re-level SES, with n emphsis on exmining those chrcteristics relted to ccess to nd utiliztion of TB services. This study rises other ctionble next steps including understnding wht fctors re tied in to disese severity, both individully nd t the community level, whether the SES-severity ssocition is further modified by other fctors such s rce, nd the potentil impct of SES on delys leding to more severe dignoses. Approvl ws grnted for this study in My 2009 from the University of Wshington nd Wshington Stte Institutionl Review Bords. The uthors thnk the Public Helth Settle & King County Tuberculosis Control Progrm for ccess to dt, nd the stff for their support of this project. REFERENCES 1. Wllce RM, Kmmerer JS, Idemrco MF, Althomsons SP, Winston CA, Nvin TR. Incresing proportions of dvnced pulmonry tuberculosis reported in the United Sttes: re delys in dignosis on the rise? Am J Respir Crit Cre Med 2009;180:1016-22. 2. Decrese in reported tuberculosis cses United Sttes, 2009. MMWR Morb Mortl Wkly Rep 2010;59(10):289-94. 3. Ntionl Tuberculosis Controllers Assocition, Centers for Disese Control nd Prevention (US). Guidelines for the investigtion of contcts of persons with infectious tuberculosis. Recommendtions from the Ntionl Tuberculosis Controllers Assocition nd CDC. MMWR Recomm Rep 2005;54(RR-15):1-47. 4. Loudon RG, Spohn SK. Cough frequency nd infectivity in ptients with pulmonry tuberculosis. Am Rev Respir Dis 1969;99:109-11. 5. Schechter MS, Mrgolis PA. Reltionship between socioeconomic sttus nd disese severity in cystic fibrosis. J Peditr 1998;132:260-4. 6. Rbin DL, Richrdson MS, Stein SR, Yeger H Jr. Srcoidosis severity nd socioeconomic sttus. Eur Respir J 2001;18:499-506. 7. Steptoe A, Hmer M, O Donnell K, Venurju S, Mrmot MG, Lhiri A. Socioeconomic sttus nd subclinicl coronry disese in the Whitehll II epidemiologicl study. PLoS One 2010;5:e8874. 8. Klssen AC, Curriero FC, Hong JH, Willims C, Kulldorff M, Meissner HI, et l. The role of re-level influences on prostte cncer grde nd stge t dignosis. Prev Med 2004;39:441-8. 9. McKinnon JA, Duncn RC, Hung Y, Lee DJ, Fleming LE, Voti L, et l. Detecting n ssocition between socioeconomic sttus nd lte stge brest cncer using sptil nlysis nd re-bsed mesures. Cncer Epidemiol Biomrkers Prev 2007;16:756-62. 10. Giunt A, Arcsoy SM, Ptel N, Wilt J, Lederer DJ. Low socioeconomic sttus is ssocited with greter disese severity in idiopthic pulmonry fibrosis. Abstrct presented t the 2010 Interntionl Conference of the Americn Thorcic Society; 2010 My 14 19; New Orlens. 11. Storl DG, Yimer S, Bjune GA. A systemtic review of dely in the dignosis nd tretment of tuberculosis. BMC Public Helth 2008;8:15. 12. Frh MG, Rygh JH, Steen TW, Selmer R, Heldl E, Bjune G, et l. Ptient nd helth cre system delys in the strt of tuberculosis tretment in rwy. BMC Infect Dis 2006;6:33. 13. Kiwuw MS, Chrles K, Hrriet MK. Ptient nd helth service dely in pulmonry tuberculosis ptients ttending referrl hospitl: cross-sectionl study. BMC Public Helth 2005;5:122. 14. Shermn LF, Fujiwr PI, Cook SV, Bzermn LB, Frieden TR. Ptient nd helth cre system delys in the dignosis nd tretment of tuberculosis. Int J Tuberc Lung Dis 1999;3:1088-95. 15. Mesfin MM, Newell JN, Wlley JD, Gessessew A, Mdeley RJ. Delyed consulttion mong pulmonry tuberculosis ptients: cross sectionl study of 10 DOTS districts of Ethiopi. BMC Public Helth 2009;9:53. 16. Ngmvithypong J, Yni H, Winkvist A, Diwn V. Helth seeking behviour nd dignosis for pulmonry tuberculosis in n HIVepidemic mountinous re of Thilnd. Int J Tuberc Lung Dis 2001;5:1013-20. 17. Benjmins MR, Kirby JB, Bond Huie SA. County chrcteristics nd rcil nd ethnic disprities in the use of preventive services. Prev Med 2004;39:704-12. 18. Pho LL, Telemn MD, Wng YT, Chee CB. Chrcteristics of ptients with delyed dignosis of infectious pulmonry tuberculosis. Respirology 2005;10:196-200.
Socioeconomic Disdvntge nd TB: 2000 2008 109 19. Rdhkrishn S, Frieden TR, Subrmni R, Snth T, Nrynn PR; Indin Council of Medicl Reserch. Additionl risk of developing TB for household members with TB cse t home t intke: 15-yer study. Int J Tuberc Lung Dis 2007;11:282-8. 20. Psipnody JG, Miller TL, Vecino M, Mungui G, Grmon R, Be S, et l. Pulmonry impirment fter tuberculosis. Chest 2007;131:1817-24. 21. Golub JE, Bur S, Cronin WA, Gnge S, Bruch N, Comstock GW, et l. Delyed tuberculosis dignosis nd tuberculosis trnsmission. Int J Tuberc Lung Dis 2006;10:24-30. 22. Helke KL, Mnkowski JL, Mnbe YC. Animl models of cvittion in pulmonry tuberculosis. Tuberculosis (Edinb) 2006;86:337-48. 23. Centers for Disese Control nd Prevention (US). CDC tuberculosis surveillnce dt trining: report of verified cse of tuberculosis (RVCT): instruction mnul. Atlnt: CDC; 2009. 24. Census Bureu (US). Census block groups crtogrphic boundry files descriptions nd metdt: 2005 [cited 2012 Oct 22]. Avilble from: URL: http://www.census.gov/geo/www/cob/bg_metdt. html 25. Krieger N, Chen JT, Wtermn PD, Soobder MJ, Subrmnin SV, Crson R. Geocoding nd monitoring of US socioeconomic inequlities in mortlity nd cncer incidence: does the choice of re-bsed mesure nd geogrphic level mtter?: the Public Helth Disprities Geocoding Project. Am J Epidemiol 2002;156:471-82. 26. Centers for Disese Control nd Prevention (US). Tuberculosis Informtion Mngement System [cited 2012 v 27]. Avilble from: URL: http://www.cdc.gov/tb/progrms/tims/tims.htm 27. Census Bureu (US). Summry file 3: 2000 census of popultion nd housing. Wshington: Census Bureu; 2007. 28. Census Bureu (US). Summry file 1: 2000 census of popultion nd housing. Wshington: Census Bureu; 2007. 29. Luke DA. Multilevel modeling. Thousnd Oks (CA): Sge Publictions Inc.; 2004. 30. Rbe-Hesketh S, Skrondl A, Pickles A. Generlized multilevel structurl eqution modeling. Psychometrik 2004;69:167-90. 31. Klssen AC, Kulldorff M, Curriero F. Geogrphicl clustering of prostte cncer grde nd stge t dignosis, before nd fter djustment for risk fctors. Int J Helth Geogr 2005;4:1. 32. World Helth Orgniztion. Addressing poverty in TB control. Options for ntionl TB control progrmmes. Genev: WHO; 2005. 33. Long NH, Johnsson E, Diwn VK, Winkvist A. Different tuberculosis in men nd women: beliefs from focus groups in Vietnm. Soc Sci Med 1999;49:815-22. 34. Asch S, Leke B, Anderson R, Gelberg L. Why do symptomtic ptients dely obtining cre for tuberculosis? Am J Respir Crit Cre Med 1998;157:1244-8. 35. Mdebo T, Lindtjorn B. Dely in tretment of pulmonry tuberculosis: n nlysis of symptom durtion mong Ethiopin ptients. Med Gen Med 1999:E6. 36. Centers for Disese Control nd Prevention (US). Helth, United Sttes, 2011, with specil feture on socioeconomic sttus nd helth [cited 2012 Oct 22]. Avilble from: URL: http://www.cdc.gov/nchs/hus/specil.htm 37. Ady LA, Andersen RM. Equity of ccess to medicl cre: conceptul nd empiricl overview. Med Cre 1981;19:4-27. 38. Dunlop S, Coyte PC, McIsc W. Socio-economic sttus nd the utilistion of physicins services: results from the Cndin Ntionl Popultion Helth Survey. Soc Sci Med 2000;51:123-33. 39. Alter DA, Nylor CD, Austin P, Tu JV. Effects of socioeconomic sttus on ccess to invsive crdic procedures nd on mortlity fter cute myocrdil infrction. N Engl J Med 1999;341:1359-67. 40. Public Helth Settle & King County. King County community helth indictors [cited 2012 Oct 18]. Avilble from: URL: http:// www.kingcounty.gov/helthservices/helth/dt/chi.spx 41. Oeltmnn JE, Kmmerer JS, Pevzner EE, Moonn PK. Tuberculosis nd substnce buse in the United Sttes, 1997 2006. Arch Intern Med 2009;169:189-97. 42. Oeltmnn JE, Oren E, Hddd MB, Lke LK, Hrrington TA, Ijz K, et l. Tuberculosis outbrek in mrijun users, Settle, Wshington, 2004. Emerg Infect Dis 2006;12:1156-9. 43. Asch S, Leke B, Knowles L, Gelberg L. Tuberculosis in homeless ptients: potentil for cse finding in public emergency deprtments. Ann Emerg Med 1998;32:144-7. 44. Altet-Gômez MN, Alcide J, Godoy P, Romero MA, Hernndez del Rey I. Clinicl nd epidemiologicl spects of smoking nd tuberculosis: study of 13,038 cses. Int J Tuberc Lung Dis 2005;9:430-6. 45. Alisjhbn B, Shirtmdj E, Nelwn E, Purw AM, Ahmd Y, Ottenhoff TH, et l. The effect of type 2 dibetes mellitus on the presenttion nd tretment response of pulmonry tuberculosis. Clin Infect Dis 2007;45:428-35. 46. Geng EH, Kreiswirth BN, Burzynski J, Schluger NW. Trnsmission trends for humn immunodeficiency virus ssocited tuberculosis in New York City. Int J Tuberc Lung Dis 2005;9:661-6. 47. Dley CL, Smll PM, Schecter GF, Schoolnik GK, McAdm RA, Jcobs WR Jr, et l. An outbrek of tuberculosis with ccelerted progression mong persons infected with the humn immunodeficiency virus. An nlysis using restriction-frgment-length polymorphisms. N Engl J Med 1992;326:231-5. 48. Crucini M, Mlen M, Bosco O, Gtti G, Serpelloni G. The impct of humn immunodeficiency virus type 1 on infectiousness of tuberculosis: met-nlysis. Clin Infect Dis 2001;33:1922-30.