Overview Chapter 6: Impact Indicatrs The best measure f the lng-term impact f all HIV preventin activities is the HIV incidence rate, namely the number f new cases f HIV infectin per year divided by the number f HIVnegative individuals in the ppulatin at the start f the year. Data n HIV incidence are scarce, hwever, and usually frm small grups, nt natinally representative samples. It is easier t measure the prevalence f HIV infectin the prprtin f the ttal ppulatin that is infected with HIV. The gal f HIV preventin prgrammes is t reduce the transmissin f HIV. Since peple aged under 25 have had a relatively shrt time in which t becme infected, mst prevalent infectins in this age-grup will have been recently acquired. Prevalence in this grup can therefre be a gd measure f the rate at which the epidemic is prgressing and can shw where preventin prgrammes are making a difference. The number f new HIV infectins that ccur amng yung peple may reflect behaviur change ccurring amngst all age grups because yung peple can be infected by lder partners. It is imprtant t be able t cnstruct a gd measure f bth the prprtin f yung peple infected with HIV and the trends in the prevalence f HIV infectin ver time. Changes in HIV prevalence amng a particular grup can ccur fr many reasns. It is almst as imprtant t be able t explain changes in prevalence as t be able t detect these changes as they ccur. This sectin defines indicatrs that describe levels f HIV infectin amng yung peple. Indicatr Tls fr Measurement Pririty Pririty Generalized Cncentrated/ Epidemic Lw level 1. HIV prevalence amng yung Sentinel surveillance pregnant wmen C A 2. HIV prevalence amng yung peple in cmmunity based surveys 3. HIV prevalence in subppulatins f yung peple with high-risk behaviur 4. Yung peple wh have a sexually transmitted infectin Natinally representative general ppulatin survey UNAIDS/WHO Secnd Generatin Surveillance Guidelines FHI guidelines n sampling in sub-ppulatins Surveys amng clinic-based grups (ANC attendees, gynaeclgy patients, bld dnrs, etc.) Cmmunity-based surveys C A A A C A
1. HIV prevalence amng yung pregnant wmen Pririty: Cre in generalised epidemics, additinal in thers Definitin Prprtin f yung pregnant wmen that test psitive fr HIV during 'unlinked annymus sentinel surveillance' at selected antenatal clinics Target Ppulatin Pregnant yung wmen, aged 15-24 years Numeratr Number f pregnant yung wmen wh test psitive fr HIV infectin using an unlinked annymus testing prtcl, while attending antenatal clinics (ANC) Denminatr All yung pregnant wmen wh are tested fr HIV infectin using an unlinked annymus testing prtcl, while attending antenatal clinics Measurement tls UNAIDS/WHO Secnd Generatin Surveillance guidelines What it measures In mst cuntries, yung wmen wh attend antenatal clinics are a reasnably representative sample f yung wmen in the general ppulatin. Yung wmen wh are pregnant have, by definitin, had unprtected sex smetime in the last ten mnths and have therefre ptentially been sexually expsed t HIV infectin. They will nt be, n average, a grup characterized by ther high-risk behaviur. Participatin bias is relatively lw in this sample because HIV testing is carried ut annymusly with bld that is rutinely taken frm all pregnant wmen fr ther rutine tests. Mst HIV infectins amng yung wmen will have been recently acquired. Trends in HIV prevalence in this grup may therefre reflect trends in the incidence f new HIV infectins. Hw t measure it HIV prevalence is estimated frm the unlinked annymus testing f bld samples that are rutinely taken frm pregnant wmen f all ages at sentinel antenatal clinics. The quality f these data depends n the structure f the surveillance system. An ideal sentinel surveillance system wuld include clinics chsen t reflect a cuntry s urban, rural, ethnic and ther scigegraphic divisins. The methds used fr surveillance shuld be the same in all sites. The indicatr shuld be reprted as a percentage and presented separately fr age in three grups: 15 19, 20 24 and 15 24. It shuld als be reprted by parity in tw grups primigravidae and multigravidae. If the sample is large enugh the results can be presented
disaggregated by bth age and parity. Parity is imprtant because prevalence amng wmen having their first pregnancy prvides a better estimate f incidence. The mean prevalence shuld be reprted, tgether with the number f clinics cntributing data, the ttal number f wmen tested, and the ttal number f wmen wh tested psitive fr HIV infectin. The data shuld be presented fr the capital city, ther urban areas, and rural areas. Strengths and limitatins In cuntries where the epidemic is hetersexually driven, this indicatr gives a fairly gd idea f relatively recent trends in HIV infectin natinwide. It is less reliable as an indicatr f verall epidemic trends in areas where the bulk f HIV infectin remains cnfined t sub-ppulatins with especially high-risk behaviur. In these circumstances it is a useful way f mnitring whether HIV infectin is spreading beynd these sub-ppulatins. T interpret changes in ANC prevalence it is imprtant t islate real changes in the prprtin f yung wmen wh are infected with HIV frm artefacts f the surveillance system. The HIV prevalence bserved amng yung wmen attending ante-natal clinics may change fr a number f reasns nt directly cnnected with the true prevalence f HIV infectin amng yung wmen in the general ppulatin. Changes that affect the number f yung wmen wh becme pregnant, the prprtin f thse wh seek antenatal care, and the stage f the pregnancy at which a wman first visits an antenatal clinic culd all affect the HIV prevalence bserved at ANC. Sme f these changes, such as an increase in the age at first sex, may als affect the incidence f new HIV infectins amng yung wmen. As a result, trends in the prevalence f HIV infectin amng yung pregnant wmen shuld be interpreted carefully. When mnitring trends, the sample cmpsitin is very imprtant. The representativeness f the clinic sample is nly as gd as the data n which the sampling frame was based. Accurate infrmatin abut the size and lcatin f the clinics means that this can be treated with mre cnfidence. Interpretatin will be easier if the same sample f clinics is used in several runds f surveillance. The size f the clinics has t change cnsiderably befre the sample will cease t be representative. If prevalence appears t be changing, it may make sense t retain the same sample lng enugh t establish whether a trend is emerging. HIV prevalence amng yung pregnant wmen can als be used t estimate HIV prevalence amng yung wmen in the general ppulatin. Sftware is available (at www.unaids.rg) t develp the adjustments necessary t make these data representative f the ppulatin. Hwever, estimatin is beynd the scpe f this guide.
2. HIV prevalence amng yung peple in cmmunity-based surveys Pririty: Cre in generalised epidemics, additinal in thers Definitin Prprtin f yung peple wh test psitive fr HIV in a general ppulatin survey Target Ppulatin 15-24 year lds Numeratr Number f yung peple wh test psitive fr HIV infectin Denminatr All yung peple tested Measurement tls Natinally, r reginally, representative cmmunity based surveys, which include cllectin f suitable bilgical specimens. What it measures Data frm cmmunity-based surveys are ptentially the best surce f data n HIV prevalence amng yung peple in the general ppulatin. Hwever, they may nt prvide gd estimates fr subgrups f the yuth ppulatin (e.g., IDUs) whse behaviur wuld place them in a grup at high risk f HIV infectin, because a cmmunity-based survey is unlikely t find sufficient peple in these categries t make a representative sample. Data fr this indicatr is nt as feasible t cllect as ANC surveillance data is, fr tw main reasns. The first is that because very few such surveys have been carried ut, there is nt enugh evidence t recmmend that these data be used as an essential part f the HIV/AIDS mnitring amng yung peple. The secnd is that surveys f this srt are expensive and time-cnsuming. T prvide rbust estimates f prevalence trends, they must be repeated at regular intervals in a cmparable manner. If such a survey can nly be carried ut at infrequent intervals, the findings can be cmpared with the results frm ANC surveillance Hw t measure it This indicatr shuld be reprted as a percentage brken dwn by sex and by age in three grups: 15-19, 20-24, and 15-24. The unweighted sample sizes and respnse rates shuld be given fr each categry. The HIV testing prtcl shuld als be given. Strengths and limitatins The findings f a general ppulatin survey can be taken at face value, prvided the survey is truly representative f the ppulatin in which it was carried ut.
General ppulatin surveys apprach participants, while mst ther methds f data cllectin rely n participants presenting themselves at the place where HIV testing is being carried ut. This means that the selectin and participatin bias shuld be less imprtant in these surveys. If the survey s sampling frame is inaccurate, hwever, r if the survey is badly implemented, there may still be sme selectin bias. Participatin bias is ptentially a greater prblem. The extent f participatin bias will be influenced by the tpic f the survey and the prtcl under which the HIV testing is carried ut. The factrs f particular cncern are thse that may relate t the HIV status f the ptential respndent (e.g., high-risk sexual behaviur). If the individuals wh chse nt t participate differ markedly frm thse wh d participate n such characteristics, the accuracy f the HIV-prevalence estimates may be affected. If basic infrmatin is cllected frm thse wh d nt take part in the survey (r in part f the survey),participatin bias can be detected, althugh nt crrected, at the analysis stage. Respnse rates fr the survey shuld always be reviewed and presented where pssible. A ptentially serius limitatin f cmmunity- based survey data is lack f cntinuity. Because surveys are expensive and time-cnsuming, the scpe and frmat f successive surveys may vary. This intrduces an unquantifiable errr int the estimates. Reliable data n HIV prevalence ver time requires a series f cmparable surveys t be carried ut peridically in the same ppulatin.
3. HIV prevalence in sub-ppulatins f yung peple with high-risk behaviur Pririty: Cre in cncentrated epidemics, additinal in thers Definitin Prprtin f yung members f defined sub-ppulatin at higher risk f cntracting r transmitting HIV infectin wh test HIV psitive Target Ppulatin Members f a high-risk grup, aged 15-24 year lds Numeratr Number f yung peple in a high-risk grup wh test psitive fr HIV infectin Denminatr Yung peple at high risk f cntracting r transmitting HIV Measurement tls UNAIDS/WHO Secnd Generatin Surveillance guidelines; FHI guidelines n sampling in subppulatins. What it measures This indicatr is mst useful in cuntries where HIV infectin has nt spread t the general ppulatin but remains cncentrated in certain grups. The prevalence f HIV infectin amng members f these grups identifies imprtant areas/grups fr interventin. Trends in prevalence can indicate whether interventins are having an impact r whether sme ther factrs are driving prevalence up r dwn. In a cncentrated epidemic, the grups f interest generally include ne r mre f the fllwing: injecting drug users, men wh have sex with ther men, sex wrkers and frequent clients f sex wrkers, and street children. Hw t measure it This indicatr shuld be reprted as a percentage brken dwn by sex and by age in three grups: 15 19, 20 24 and 15 24. Any data available n yung peple f 10 t 14 years can als be given. The sample sizes shuld be given fr each categry. The HIV testing prtcl used shuld be given. It may be apprpriate t give estimates disaggregated by the duratin f risk-grup membership. Surveys cnducted amng high-risk grups shuld nt sample nly yung peple; instead, this indicatr shuld be based n the data frm a subset f respndents. It is imprtant that surveys amng high-risk grups cver a sufficiently large sample t prvide reliable estimates fr yung peple.
If sample sizes are small and such subdivisins wuld prejudice annymity, r if the infrmatin is nt available in relatin t HIV status, it is nt necessary t prvide the prevalence data subdivided by age r duratin f grup membership. Instead, the age distributin f whle grup shuld be reprted, regardless f HIV status. The grup can be described in three age grups: <15, 15 19, 20 24. Where available, the median duratin f risk-grup membership shuld be reprted fr each age-grup. Tracking HIV in sub-ppulatins can be lgistically and ethically difficult, especially if the grups are marginalized r their activities are illegal. Sampling and estimatin f ttal ppulatin sizes are key issues. An understanding f hw the sampled ppulatin relates t any larger ppulatin sharing similar risk behaviur is critical t the interpretatin f the indicatr. Fr sme grups, ppulatin-based sampling strategies will be necessary. In ther cases, sentinel sites are available. Sentinel sites fr these ppulatins tend t be linked t the prvisin f health services fr example, a men s health clinic in an area with a high cncentratin f gay sex bars, r a drug rehabilitatin centre. Strengths and limitatins A limitatin f surveys amng high-risk grups is that it is nt usually pssible t find a representative prbability sample. This indicatr will, at best, represent the members f the sub-grup with high-risk behaviur frm which the sample was drawn, and may nt represent all persns displaying that behaviur. This means that it is difficult t estimate the extent t which an indicatr based n these data will describe prevalence amng all members f the grup. Infrmatin n the size f the high-risk grups is necessary t put this prevalence data int a natinal (r reginal) cntext. Because f the difficulties in access t sub-ppulatins, the biases in sub-ppulatin sersurveillance data are likely t be far greater (and much less predictable) than thse in data frm a mre generalized ppulatin, such as wmen at antenatal clinics. Where sentinel sites prvide health services t the sub-ppulatin in questin fr example, the use f the facility may be assciated with prblems that are themselves related t HIV infectin. It is especially difficult t minimize biases assciated with age, since the age f participatin in especially high-risk behaviur may be very variable. Chrnlgical age is less imprtant in a high-risk grup than duratin f membership f the grup when it cmes t explaining bserved patterns. It is essential, hwever, t cllect and present data by age because it is this infrmatin which allws the targeting f interventins and plicies. Changes in HIV prevalence in these grups culd reflect the success r failure f preventin attempts but they may als reflect change in recruitment and expsure, which are unrelated t preventin effrts. This indicatr shuld be cnsidered in cnjunctin with the behaviural indicatrs that refer t membership and activity f high-risk grups, because changes in recruitment t, r exit frm, the grup culd be respnsible fr change in bserved prevalence. Prevalence will als be affected by changes in the number f new infectins and in mrtality. Despite these difficulties, it is essential t track HIV infectin in thse with higher-risk behaviur in cncentrated epidemics. The infrmatin will nt be perfect, but sme measure f prgress r f lack f prgress will be essential t maintain supprt fr preventin prgrammes in critical sub-ppulatins
4. Yung peple wh have a sexually transmitted infectin Pririty: Additinal Definitin Prprtin f diagnstic tests fr an STI cnducted in yung peple that cnfirm the presence f an STI Target Ppulatin 15-24 years Numeratr Number f diagnstic tests carried ut fr patients aged 15-24 years which cnfirm the existence f an STI Denminatr Ttal number f diagnstic tests carried ut fr patients aged 15-24 years * *NOTE: The type(s) f STI shuld be dependent n what is lcally imprtant. If mre than ne infectin is cnsidered, the results shuld be given separately. What it measures In cuntries where HIV infectin has yet t take hld, this indicatr gives sme idea f the ptential fr HIV t spread. In ppulatins where HIV infectin is established, STI prevalence can be used as an additinal way t mnitr the levels f risky sexual behaviur. Once HIV infectin is established in a ppulatin it is mre difficult t generalize abut the interpretatin f this indicatr. This is primarily because the epidemilgy f STIs d nt necessarily fllw that f HIV, since mst STIs have a shrter duratin f infectin, many are transmitted mre easily than HIV, and mst can be treated and cured. 1 If STIs are t be used as a marker fr risk behaviur, additinal infrmatin is needed n the incidence f new infectins and n treatment.. In cuntries where health facilities are universally available, the number f new STI diagnses in a defined regin can be used as a prxy measure f incidence. This is mst useful fr mnitring trends, because the abslute number depends n the number f undetected infectins in the ppulatin, but trends in new diagnses culd reflect changing rates f infectin. Hw t measure it 1 http://www.dh.gv.za/dcs/reprts/2000/hivreprt.html
This indicatr shuld be given as a percentage and presented separately by sex in three age grups: 15 19, 20 24 and 15 24. If data are available fr yung peple aged 10 t 14 years, these shuld als be presented. The prevalence f an STI must be determined frm labratry diagnses f infectin, because many STIs are asymptmatic and many STIs have similar symptms, which makes an accurate clinical diagnsis difficult. It is recgnized that in many cuntries, the availability f suitable labratry facilities is limited and that this may restrict the surces f data fr this indicatr. This indicatr places ne limit n the surce f the data. Data shuld nt be cllected frm STIclinic patients, r frm any ther grup fr which a test is being carried ut because the persn has STI symptms. This is because this grup is very highly selected and the prevalence f infectin amngst the grup des nt prvide infrmatin n the prevalence f infectin amngst yung peple in the general ppulatin. Suitable grups r surces f grups include yung wmen attending ANC, gynaeclgy patients, family-planning-clinic screening prgrammes, ccupatinal screening prgrammes, bld dnrs r frm cmmunity-based surveys. The interpretatin f this indicatr will depend n the surce f the data, and the surce must therefre be reprted with the indicatr estimates. Strengths and limitatins This indicatr des nt necessarily give infrmatin n hw many peple are infected with STI. Using the number f requests fr diagnstic tests in the denminatr, instead f peple, means that the results will be difficult t generalize t the ppulatin, if the surce is rutine clinicbased data. If the data are frm screening prgrammes r cmmunity surveys it will be easier t relate this indicatr t STI prevalence because each test will crrespnd t ne persn. Selectin and participatin bias are likely t be prblems if the indicatr is based n data frm health services: thse attending are likely t be systematically different frm thse wh d nt attend and thse wh cnsent t an STI test are als likely t be different frm thse wh d nt cnsent. It is impractical t limit this indicatr t thse STI that can be diagnsed frm samples taken fr ther purpses, as sme f the mre imprtant STI cannt be diagnsed in this way. If the data are derived frm a screening prgramme r survey, fr which peple are sught fr testing rather than vice versa, selectin bias is less f a prblem. If the survey uses unlinked annymus testing, participatin bias will be minimized. As a screening prgramme cannt by its nature be annymus, participatin bias must be cnsidered.