Lecture Contents Frequency –Prevalence –Incidence Cumulative Density Precision –P value –Confidence level Association - Relative risk - Rate ratio - Risk ratio - Odds ratio - Risk difference
Frequency Measures Two types: –Someone has the disease already: PREVALENCE –Someone gets the disease in the future: INCIDENCE
Study Design Direction of inquiry Cohort Case-control Historical cohort Survey / Cross Sectional TODAY
Prospective Cohort Start here ** * toto t1t1 Free of outcome ExposureOutcome
Historical Cohort Start here ** * toto t1t1 Free of outcome ExposureOutcome
Case Control Start here Case Control + Popu lation Exposure Outcome
RASIO DAN PROPORSI RASIO PERBANDINGAN SECARA UMUM TAK ADA KAITAN PEMBILANG DAN PENYEBUT PROPORSI PEMBILANG MERUPAKAN BAGIAN DARI PENYEBUT A/B A/(A+B)
Frequency measures: diagnostic research Suppose: you see a patient with symptoms that possibly point at arthritis Research question ?
Frequency measures: diagnostic research Suppose: you see a patient with symptoms that possibly point at venous thrombosis Research question: What is the probability of arthritis given the physical exam / tests?
Frequency measures: prevalence Cross-sectional studies –Determinant and disease measured at the same time Prevalence –Number of persons with the disease at a certain moment
Frequency measures: prevalence Prevalence (%) = Number of persons with the disease Total population Numerator is part of denominator
ANGKA POINT PREVALENSI MERUPAKAN NILAI PROPORSI PADA SATU SAAT TERTENTU GUNA EVALUASI PENGOBATAN KASUS (BARU+LAMA) SATU SAAT SELURUH POPULASI SAAT ITU
POINT PREVALENCE Tujuan : mengetahui prevalensi artritis di suatu komunitas di suatu hari tertentu Hari itu kita lakukan kunjungan dari rumah ke rumah untuk melakukan anamnesis dan pemeriksaan fisik untuk menentukan berapa orang yang mengalami artritis pada hari itu Prevalensi (point) = Jumlah orang yang mengalami artritis hari itu Jumlah penduduk di komunitas hari itu
PERIOD PREVALENCE BARU+LAMA SUATU PERIODE SELURUH POPULASI PERIODE TERSEBUT
Frequency measures: prevalence Examples –50% of the persons with a suspicion of lung cancer had a lesion on the thorax X- ray –In a general practice population of 2500 persons, 50 had asthma –30% of the Dutch people smoke
Frequency measures: prognostic research Suppose: You see a patient diagnosed as MCI post CABG who asks for her prognosis Research question?
Frequency measures: prognostic research Suppose: You see a patient diagnosed as MCI post CABG who asks for her prognosis Research question: What is the probability that I die within 5 years / get a relapse?
ANGKA INSIDENSI POPULASI RENTAN = BEBAS KASUS MENURUT PERIODE WAKTU PENYEBUT = POPULASI RENTAN (memiliki kemungkinan untuk menjadi kasus) EVALUASI PENCEGAHAN KASUS BARU DLM SUATU PERIODE POPULASI RENTAN DLM PERIODE TERSEBUT
Incidence per 1000 = Incidence per = KASUS BARU DLM SUATU PERIODE X 1000 POPULASI RENTAN DLM PERIODE TERSEBUT KASUS BARU DLM SUATU PERIODE X POPULASI RENTAN DLM PERIODE TERSEBUT
Frequency measures: Incidence Incidence –Number of new cases –In the population at risk Two types of incidence –Cumulative Incidence Risk (CIR) –Incidence Density Rate (IDR)
Cumulative Incidence Risk (CIR) calculation Outcome (+)(-) Exposurea b Non Exposurecd a /(a +b) = CIR outcome in expose c /(c +d) = CIR outcome in non expose
Post CABGOutcome DeadAlive Anterior/Inferior MCI8 21 Non Anterior/Inferior MCI 1297 Calculate : CIR death post CABG MCI anterior/inferior ?CIR death post CABG MCI anterior/inferior ? CIR death pada post CABG non MCI ant./inf ?CIR death pada post CABG non MCI ant./inf ?
CIR death post CABG MCI anterior/inferior ?CIR death post CABG MCI anterior/inferior ? 8/29 = = 27.59% CIR death pada post CABG non MCI ant./inf ?CIR death pada post CABG non MCI ant./inf ? 12/109 = = 11.01% 12/109 = = 11.01% Post CABGOutcome DeadAlive Anterior/Inferior MCI8 21 Non Anterior/Inferior MCI 1297
Frequency measures: Incidence Cumulative incidence –new cases in a certain time period in the population at risk (free of the disease/outcome at the start) –proportion / probability –varies between 0 and 1 –within certain time period
Frequency measures: Incidence Cumulative incidence: examples –5-year risk of a second MI –10-year survival for women with breast cancer –1-year risk of a fracture for osteoporotic women
Frequency measures Incidence Incidence Density = number new patients person-years of the population at risk 10 per 1000 person-years (PY) between 0 and infinity
Incidence Density Rate (IDR) calculation Outco me (+) Person- time Exposurea t 1 Non Exposurect2t2 a /(t 1) = IDR outcome in expose c /(t 2) = IDR outcome in non expose
Incidence Density Rate (IDR) calculation 35 WEEKS Relaps (+) Person-time Non Radiotherapy Radiotherapy9359 a /(t 1) = IDR relaps in non radiotherapy ? c /(t 2) = IDR relaps in radiotherapy ?
RESULT After 35 weeks follow up, we have 9 events of new cell growth in Ca cerviks stage 2 with radiotherapy out of 359 person-week, giving an incidence rate of new cell growth in patient Ca cervix stage 2 with radiotherapy is : 9 / 359 = cases / 359 person-weeks = 25 cases / 1000 person in 35 weeks or 37 cases / 1000 person in a year
RESULT After 35 weeks follow up, we have 21 events of new cell growth in Ca cerviks stage 2 without having radiotherapy out of 182 person-week, giving an incidence rate of new cell growth in patient Ca cerviks stage 2 without having radiotherapy is 21 / 182 = cases / 182 person-weeks = 115 cases / 1000 person in 35 weeks or 172 cases / 1000 person in a year
INSIDENCE DAN PREVALENCE
HUBUNGAN NILAI P = I x d Prevalens Insidens
Ad question 1: tonsillitis A.Dutch population B.1 year C.incidence D.19/1000 or 1.9%
Exercise 1 Ad question 2: asthma A.Children in the general practice B.Certain moment (look into practice data at a certain moment) C.Prevalence
Exercise 1 Ad question 3: breast cancer A.Women B.Life C.Incidence
Exercise 1 Ad question 4: vertebral collapse A.9% B year-old men and women C.Certain moment D.Prevalence
Exercise 1 Ad question 5: fractures A.Post-menopausal women B.Follow-up duration of the study C.Incidence
Frequency measures: Incidence How do we calculate a cumulative incidence?
Frequency measures: example cohort 13 persons followed for 5 years for mortality –A x--Moves away –B x Death –C breast cancer/death –D x alive –E x lost to follow-up –F x alive –G x breast cancer/death –H x-Myocardial infarction/death –I death –J x alive –K lost to follow-up –L x moves from the area –M x alive
Frequency measures: example cohort CI = 5/13 = 38% Incidence density ?
Frequency measures: Etiologic research Suppose: you see a patient with lung cancer, who asks for the possible cause Research question?
Frequency measures: Etiologic research Suppose: you see a patient with lung cancer, who asks for the possible cause Research question: Is smoking a risk factor for lung cancer?
Measures of association Epidemiology –Disease = f (determinants) –Is the determinant associated with the disease? –Is the probability of disease different for exposed and non-exposed Ratio risk :Outcome risk in exposed Outcome risk in non exposed CIR ratio or IDR ratio
Measures of association: Cohort approach Research question? Is smoking associated with lung cancer? Cohort approach –divide the cohort in smokers and non-smokers –estimate the incidence density (or CI) in each group –prior: ID smokers > ID not smokers
Association measurement RELATIVE RISK (RR) = Outcome incidence in expose group Outcome incidence in non expose group –RISK RATIO (CIR ratio) –RATE RATIO (IDR ratio)
(CIR) Outcome Risk Ratio (+)(-) Exposeda b Non exposedcd1,00 Risk Ratio = a /(a +b) c /(c +d)
Post CABGOutcome DeadAlive Anterior/Inferior MCI8 21 Non Anterior/Inferior MCI 1297 Calculate : Risk of death in post CABG ant/inf MCI compare to non ant/inf MCI ?
Hasiljadi RR (95% CI) MatiHidup Anterior/inferior8 21 Bukan ant./inferior12971,00 Risk Ratio = a /(a +b) c /(c +d) Interpretation ? 2.51 ( )
Hasiljadi RR (95% CI) MatiHidup Anterior/inferior8 21 2,51 (1,13-5,55) Bukan ant./inferior12971,00 Kesimpulan: Pasien IMA anterior / inferior secara signifikan mempunyai risiko mati 2,5 kali lipat lebih tinggi jika dibandingkan dengan pasien penderita IMA bukan anterior/inferior, post CABG di ICCU.
Interpretasi hasil : RR (OR) < 1 = exposure as protective factor for outcome occurance RR (OR) = 1 = No occurance difference between exposed and non exposed RR (OR) > 1 = exposure as risk factor for outcome occurance
Rumus dasar (IDR) Rate Ratio Outcom e (+) Person- time Exposeda t 1 Non exposedct2t2 1,00 Rate Ratio = a /(t 1) c /(t 2)
Measures of association: Cohort approach Smoking and lung cancer Lung cancer YesNo Rokok py Determinant Tidak Rokok py Hitung:Risiko terjadinya Ca PARU pada perokok dibandingkan dengan tidak perokok?Hitung:Risiko terjadinya Ca PARU pada perokok dibandingkan dengan tidak perokok?
Measures of association: Cohort approach Smoking and lung cancer Disease YesNo Yes py Determinant No py RR = (440/22.008) / (212/21.235) = 2.0
Measures of association Risk difference (RD) between exposed and non- exposed reflects public health impact = CIR exposed – CIR nonexposed or = IDR exposed - IDR nonexposed Risk difference smoking and lung cancer –RD = 20/1000 py - 10/1000 py = 10 / 1000 py
Measures of association: Case Control approach Research question: Does smoking increase the risk of lung cancer ? Patient control study –select cases and controls –Estimate the frequency of smoking among cases and controls –prior: % smokers among cases > % smokers among controls
Measures of association: Case Control approach Disease YesNo Yes ab Determinant No cd RR? Odds ratio = (a/c) / (b/d) = ad / bc
Measures of association: Case Control approach Smoking and lung cancer (controls = 10% random sampling from cohort) Disease YesNo Yes Determinant No Odds ratio(440/212) / (300/350) = 2.42 RR = (440/740) / (212/562) = 1.57 (shouldn’t be calculated)
Measures of association Smoking and lung cancer Disease Yes No Yes Determinant No Now entire cohort as control RR = (440/3440) / (212/3712) = 2.23 Odds ratio =(440/212) / (3000/3500) = 2.42 RR (a/(a+b)) / (c/(c+d)) ~ (a/c) / (b/d)
Frequency measures: Therapeutic research Suppose: you see a patient with an increased blood pressure who you want to treat with blood pressure decreasing drugs. He asks about the effect of this treatment on the prognosis Research question: Does treatment decrease the probability of CVD?
Frequency measures: Incidence Intervention study (RCT) –Estimate incidence density (or CI) for each group –prior: ID treated < ID not treated
Exercises 2 and 3
Exercise 2 A.People of age 55 years and older B.5 years C.Incidence (probably cumulative) D.Relative risk and risk difference
Exercise 2 Risk smokers = 41/1736 = Risk non-smokers = 107/5949 = RR = 0.024/0.018 = 1.3 Smokers have a 1.3 x higher probability of CVD than non-smokers -RD = = Smokers have a 5-year risk of CVD that is 0.6% higher than that of non-smokers
Exercise 3 1.Case-control study 2.Severe head injury 3.Population 4.Alzheimer’s disease 5.Odds ratio
Exercise 3 Severe head injury in the past Alzheimer YesNo Severe Yes Head injuryNo OR = (33x167)/(31x165)=1.1
Summary Frequency and measures of association Frequency –Prevalence –Incidence cumulative density Association - Relative risk - Rate ratio - Risk ratio - Odds ratio - Risk difference
How to handle measures of association? Some other concepts
INFERENCE populasi statistik XSPXSP inferens parameter sampel
Validity and Reliability Neither Valid nor Reliable Reliable but not Valid Valid & Reliable Fairly Valid but not very Reliable Think in terms of ‘the purpose of tests’ and the ‘consistency’ with which the purpose is fulfilled/met
Reliability Do not use –Layman’s concept (someone without professional training in the subject area can understand, so that they may comprehend the issue to some degree) –Character of persons –Someone is/ is not reliable What do we use?
Outcome measures Diagnostics –Prevalence, Se, Sp, PV+, PV-, OR, Prognostics –Incidence, RR, Etiology –Incidence, RR, OR Intervention –Incidence (abs. risk), RR, RD, mean difference, NNT
Effect estimate Does a single effect estimate, e.g. RR=1.5 or RR=1.0 give sufficient information?
Effect estimate No, because it does not tell anything about precision
What kind of information do P and CI provide? No information about the validity of the study! Then what?
Dari nilai sampel, kita dapat : –Mengestimasi nilai populasi (confidence interval) –Menggeneralisir nilai sampel terhadap keadaan di populasi pengujian hipotesis Berdasarkan peluang untuk memperoleh hubungan tersebut secara kebetulan. (p value) Semakin kecil peluang adanya kebetulan, semakin besar keyakinan bahwa hubungan itu memang ada.
Hipotesa Ho : menyatakan tidak ada hubungan Ha : menyatakan ada hubungan Contoh : Is smoking a risk factor for lung cancer? Ho : Rokok bukan faktor risiko ca paru Ha : Rokok adalah faktor risiko ca paru
P value Dari uji statistik didapatkan nilai p ( probability ) P-value: besar kemungkinan hasil yang didapat/adanya hubungan hanya akibat kebetulan (often with arbitrary cut-off of 5% 0.05)
P-values Statistical significance (is not the same as clinical relevance) Dependent on –Size of the effect –Size of the study population
P-values Nilai p ini dibandingkan dengan alpha yang ditetapkan sebelumnya (often with arbitrary cut-off of 5% 0.05) Bila : P < alpha Ho ditolak p > alpha Ho diterima
Example American study on losing weight in obese people Intervention: 1.Half an hour per day sports+ diet advice 2.only half an hour sports Numbers: 2 x people
Example BMI before –Group 1:30.0 –Group 2:30.0 P < Effect size turned out to be in group 1:27.6 in group 2:27.8
Example Similar study in England Now with 2 x 50 people BMI before –Group 1:28.5 –Group 2:28.4 Weight after –Group 1:23.5 –Group 2:25.5p=0.15
P-values Paradoxal results possible: 1.Significant effect, but clinically not relevant 2.Clinically relevant effect, but not significant
ESTIMASI Walaupun kita hanya mengambil sampel, sebenarnya kita ingin mengetahui nilai populasi CLT nilai sampel = populasi, bila sampel diambil berulang kali Kenyataan sehari-hari tidak memungkinkan pengambilan sampel berulang kali Memperkirakan nilai populasi dengan nilai sampel
Confidence interval: Range of possible effect estimates that you would find if you would repeat the research (infinitely) often Objective impression of the size of the effect and the precision of the effect estimate
ESTIMASI POPULASI Point estimation Konsep deterministik Interval estimation Konsep probabilistik
ESTIMASI INTERVAL Menentukan nilai minimum dan maksimum di populasi Confidence Interval Ditentukan dengan persentase 99 %, 95 %, 90 %
Relation p-values and CI For OR and RR; if the 95%CI does not contain 1 than p < 0.05 For mean difference; if 0 not in the 95%CI than p < 0.05 And vice versa
Concluding Never consider p-values alone, but also effect estimates Present effect estimates always with confidence intervals