Anda di halaman 1dari 71

EPIDEMIOLOGI (2)

Alasan penelitian epidemiologi

Pemantauan terhadap pencemaran di

lingkungan yang meningkat (kual.&kuan)

belum diketahui efek thd kesehatan

Zat pencemar agent potensial krn : korosif, eksplosif, mutagenik, dll

Tujuan dan Ruang Lingkup

1.

Deskripsi penyakit agent, host, lingkungan

2.

Mekanisme penyakit

3.

Faktor-faktor determinan suatu penyakit

4.

Mencari data diagnostik yang spesifik

5.

Mencari cara pencegahan, pengendalian, &

pemberantasan penyakit

6.

Mengikuti berbagai faktor sbg agent potensial, identifikasi efek potensial agent

Latar Belakang Etika

Deklarasi Helsinki yang diadopsi oleh World Medical assembly ke 18 th. 1964 World Medical assembly ke 29 th. 1975

Panduan internasional utk penelitian biomedis. Indonesia diatur ol. Bagian sospol tiap daerah

Isi Panduan :

Perizinan

Penggunaan kelompok atau prosedur kontrol

Uji coba alat atau lain-lain yang belum boleh digunakan pada

manusia

Hal-hal yang perlu diperhatikan

Keadaan pelayanan kesehatan masyarakat

Harapan masyarakat adanya perbaikan penelitian jangka panjang

Unit pelayanan setempat merupakan mitra yang sangat baik

Partisipasi masyarakat setempat perencanaan, pelaksanaan, evaluasi data

Kelompok kontrol atau placebo

Pemanfaatan statistik medis

Anonimitas masyarakat untuk hal-hal yang sensitif

Penelitian perilaku

Dampak lingkungan

Etika dalam pelatihan

Model Dasar Penelitian Epidemiologi

(dapat dilakukan) di laboratorium atau

lapangan

(dapat bersifat) observasional atau experimental

Penelitian Experimental

(dapat dilakukan) terhadap hewan atau manusia

(dapat dilakukan) dalam situasi alami atau disimulasi

Membandingkan kelompok manipulasi dan non manipulasi

Lebih nyata jika melihat efek yang ditimbulkan

Dilakukan secara bertahap

Tidak selalu dapat dilakukan mengingat etika

Interpretasi hasil penelitian terbatas

Penelitian observasional

Dilakukan tanpa melakukan sesuatu terhadap

populasi secara sengaja realistis banyak faktor

yang tdk terkontrol

Desain penelitian harus baik

Metoda menghilangkan mencegah faktor

pengganggu

Dapat berupa survey jangka pendek/panjang

Dapat bersifat deskriptif kasus kendali, kohort,

cross sectional

Tabel 1. Desain dan unit penelitian epidemiologis

Unit penelitian

Desain penelitian

Nama lain

 

Penelitian

Deskriptif/analitik

Observasional

Populasi

Ekologis Cross Sectional Kasus kendali Kohort

Korelasional Prevalensi Kasus-referensi Longitudinal/follow up

Individual

Individual

Individual

 

Penelitian experimental

Studi intervensi

Penderita Orang sehat Masyarakat

 

Percobaan klinis Survey lapangan Percobaan lapangan

Keuntungan dari beberapa model penelitian observasional

 

Ekologis

Cross-

Kasus

Cohort

sectional

kendali

Penelitian penyakit

++++

-

++++

-

langka

Penelitian penyakit dengan penyebab langka

++

-

-

++++

Uji efek multipel

+

++

-

++++

penyebab

Pengukuran hubungan waktu

++

-

-

++++

Pengukuran langsung insidensi

-

-

-

++++

++++

Kecocokan

-

Tidak cocok

Model kasus kendali

Disebut case control, case history, dan retrospektif

Penyakit sudah ada/manifes tapi tidak diketahui sebabnya

Mencari penyebab penyakit yang belum diketahui

Seleksi kasus usahakan kasus baru menghindari bias

Perlu kelompok pembanding dengan keadaan yang setara

Kelompok kontrol kalau bisa komparabel dan berasal dari populasi

yang sama

Baik kasus dan kontrol tidak boleh diseleksi pertimbangan paparan terhadap agent sama

Tabel 2. Matriks 2 x 2

Tidak Sakit X

(Ď)

Paparan/Kondisi Sakit

Sakit X (D)

Terpapar faktor (E)

a

b

Tidak terpapar faktor (Ē)

c

d

2 x 2 Tidak Sakit X ( Ď) Paparan/Kondisi Sakit Sakit X (D) Terpapar faktor (E)

Penting menentukan mulai terjadinya sakit dan lamanya menderita sakit

Kesulitan kualitas dan kuantitas atas dasar ingatan

orang

Diarahkan untuk mencari perbedaan antara populasi

yang sakit dan tidak sakit sehubungan dengan penyebab

potensial

Kasus Kendali:

Waktu

Kasus Kendali: Waktu Arah penelitian Terpajan 41 Tidak terpajan 5 46 Kasus (orang sakit) Terpajan 0

Arah penelitian

Terpajan 41

Tidak terpajan 5

Kendali: Waktu Arah penelitian Terpajan 41 Tidak terpajan 5 46 Kasus (orang sakit) Terpajan 0  
Kendali: Waktu Arah penelitian Terpajan 41 Tidak terpajan 5 46 Kasus (orang sakit) Terpajan 0  

46 Kasus (orang sakit)

Terpajan 0

 
  300 kontrol (orang

300 kontrol (orang

 

tdk sakit)

Tidak terpajan 300

Tidak terpajan 300

Tabel 3. Matriks hubungan konsumsi daging dengan

enteritis necroticans

 

Enteritis necroticans

Jumlah

Ya

Tidak

Mengkon-

Ya

50

16

66

sumsi

Daging

Tidak

11

41

52

Jumlah

61

57

118

Tabel 4. Distribusi 1465 kasus kanker paru-paru dan populasi kontrol atas

dasar jenis kelamin dan konsumsi rokok 10 th sebelum terjadi kanker

Konsumsi

Laki-laki

Perempuan

sigaret

Kasus

Kontrol

Kasus

Kontrol

0

7

61

40

59

1-4

55

129

16

25

5-14

489

570

24

18

15-24

475

431

14

6

25-49

293

154

14

0

50>

38

12

0

0

Total

1357

1357

108

108

 

Bayi

Bayi Sehat

Phocomelia

Konsumsi

41

0

Thalidomide

Tidak

5

300

mengkonsumsi

Thalidomide

Keuntungan dan kerugian

Keuntungan

Dapat cepat selesai

Waktu pendek biaya murah

Informasi mudah didapat dari penderita/keluarga

Kerugian

Data berdasarkan ingatan orang kurang atau tidak akurat

Populasi kasus hanya yang masih hidup bias

Kontrol tidak berasal dari populasi yang sama bias

Model Kohort atau Prospektif

Follow up atau studi insidensi

Dilakukan ketika penyakit belum didapat tetapi sudah diketahui

adanya agent potensial

Variabel berupa agent potensial yang didefinisikan dan diukur

Kohort kesesuaian penyakit tersebut dgn yang didefinisikan pada awal

penelitian

Prospektif penelusuran dan pemantauan aksi agent thd populasi

Perbedaan dgn retrospektif pemaparan thd variabel

lingkungan yang diukur

Data bisa matriks 2 x 2

Menekankan pada efek paparan terhadap

terjadinya penyakit

Digunakan insidensi dpt dihitung resiko atribut dan resiko relatif

Tabel 5. Matriks studi Kohort Merokok dan Kanker Paru-paru

Status sakit/ Paparan

Terpapar/ Merokok

Tidak terpapar/

(E)

tidak merokok (Ē)

Sakit Ca paru-paru (D)

a

b

Tidak sakit Ca

   

paru-paru (Ď)

c

d

Model Cohort

Populasi

Model Cohort Populasi Waktu Arah penelitian   terpajan O r a n g sehat Tidak terpajan

Waktu

Model Cohort Populasi Waktu Arah penelitian   terpajan O r a n g sehat Tidak terpajan

Arah penelitian

 

terpajan 

Orang

sehat

Tidaksehat

terpajan

Waktu Arah penelitian   terpajan O r a n g sehat Tidak terpajan sakit Tidak sakit
Waktu Arah penelitian   terpajan O r a n g sehat Tidak terpajan sakit Tidak sakit

sakit

Tidak sakit

sakit

Tidak sakit

Keuntungan dan kerugian

Keuntungan

Dapat dikuantifikasi dengan akurat

Penyakit yang terjadi diperiksa dan didiagnosa dgn akurat

Tidak bias

Hubungan sebab-akibat lebih jelas/pasti

Pengukuran resiko yang sangat langsung

Kerugian

Waktu follow up lama mahal

Populasi yg tdk tetap pada lingkungan terpapar sulit memperkirakan paparan individual

Populasi pindah/meninggal sulit diganti;data sedikit

Jika penyakit jarang didapat waktu lama; drop out data selama

penelitian penelitian kohort dilakukan setelah penelitian retrospektif

Model Cross-Sectional

Studi prevalensi yg diukur prevalensi

Baik agent atau penyakit diteliti pada saat yang sama

Keadaan lingk. stabil kadar agent sama pada masa dulu dan sekarang

Lebih mudah, cepat, dan murah

Sulit menghub. antara faktor pemapar dgn prevalensi yg didapat

Berguna utk studi faktor yg bersifat permanen mis.

bangsa, gol darah

Tabel 6 Matriks Hubungan Usia Ibu Hamil dan Berat bayi

Usia Ibu

Barat bayi < 2500 gr

Berat bayi >2500 gr

Total

<20

10=n 11

40=n 12

50=n 1.

>20

15=n 21

135=n 22

150=n 2.

total

25=n .1

175=n .2

200=n…

Studi Deskriptif

Biasanya menggunakan data yg telah ada

(data sekunder)untuk menggambarkan

keadaan atau status kesehatan (angka kematian, jenis kelamin, dll)

Di Indonesia pengambilan rutin survei

rumah tangga

Berguna utk usulan penelitian epidemiologis

Contoh penelitian deskriptif

Angka kematian

bayi per 100000

lahir hidup

1000

800

600

400

200

0

1750 1800 1850 1900 1950 2000
1750
1800
1850
1900
1950
2000

Tahun

Studi Ekologis

Studi awal dengan seluruh populasi sebagai

unit

contoh: menghubungkan konsumsi garam dgn kanker oesophagus di Cina

Kesulitan menjelaskan hubungan penyebab

dan akibat

Pengolahan Data

Mencari hubungan antara agent potensial

dengan penyakit yang diteliti:

Apakah hubungan kedua variabel bermakna secara

statistik menghitung signifikansi

Hubungan berarti menghitung asosiasi:

- Resiko relatif

- Odds ratio

Menghitung resiko attribut

Menelaah hubungan kausasi

Menghitung Signifikansi

Signifikasi dlm matriks 2x2 2

2

n

n n

11

22

n n

12

21

1 n  2
1
n
2

2

n n n n

1.

2.

.1

.2

200

10*135

40*15

1 200  2
1
200
2

2

50*150*25*175

2,58

Dari tabel 2 dgn 1 derajat kebebasan & =5% nilai kritis =

3,84.

Kesimpulan hubungan usia ibu dgn berat bayi tidak bermakna tapi jika jml populasi secara proporsional 2 kali lipatnya maka hub. Kedua variabel bermakna

2

400 20*270

80*30

1 400  2 2
1
400
 2
2

100*300*50*300

5.97

Risiko Relatif:

Perbandingan antara risiko kejadian penyakit diantara yang terpajan dengan risiko kejadian penyakit diantara yang tidak

terpajan

dengan risiko kejadian penyakit diantara yang tidak terpajan Menunjukkan kekuatan hubungan kausasi Semakin besar nilainya

Menunjukkan kekuatan hubungan kausasi

Semakin besar nilainya hubungan agent dengan kejadian penyakit semakin kuat.

Data yang digunakan: Data Insidensi

Jika bukan data insidensi tetapi data prevalensi

Odd Ratio

RR =

Insidensi yg terpapar

insidensi yg tdk terpapar

RR

a

/(

a

b

)

c

/(

c

d

)

/( a  b ) c /( c  d ) • RR dapat diperoleh dari

RR dapat diperoleh dari penelitian Kohort atau kasus kendali kohort lebih pasti

Pada kasus kendali, jika tidak dapat dikaitkan dengan populasi

dihitung Odd Ratio

Odd Ratio/OR

Ukuran asosiasi yg sangat dekat dengan RR. RR=OR jika penyakit yang diteliti langka, < 20 %

Menggunakan data prevalensi

RR

OR

a

/(

a

b

)

c

/(

c

d

)

Jika penyakit langka: (a+b)≈b dan (c+d)≈d, maka:

RR OR

a / b c / d

ad

bc

Pada studi cross sectional:

p= probabilitas kondisional, kemungkinan bayi lahir kecil bila usia ibu muda= n11/n

q= kemungkinan bayi lahir normal pada usia ibu muda = n12/n

OR=p/q= (10/200)/(40/200)= 0,05/0,2 = 0,25

ibu muda = n12/n OR=p/q= (10/200)/(40/200)= 0,05/0,2 = 0,25 Resiko mendapat bayi kecil bila usia ibu

Resiko mendapat bayi kecil bila usia ibu

muda dibanding usia ibu tidak muda

Menghitung Resiko Atribut (AR/)

Karena RR bisa memberikan hasil atau angka yang sama dengan arti atau implikasi yang berbeda, maka ada perhitungan resiko atribut yg menyatakan perbedaan kedua

resiko, yaitu resiko yg terpapar dikurangi dengan resiko tidak

terpapar

p

p

δ=p -p

1

2

dimana,

resiko bagi yang terpapar
1

resiko bagi yang
2

tidak terpapar

Hubungan antara AR dan RR:

AR = (RR-1)p

AR = (OR-1)p

2

2

dan

Karena kasus kendali hanya dapat menghitung OR, maka p 2 harus didapat dari penelitian baseline yg lain

Penelitian prospektif

Insidensi Ca diantara perokok =

a/(a+c)*1000= A per 1000

Insidensi Ca diantara non perokok = b/(b+d)*1000= B per 1000

Maka :

Resiko Relatif (RR) menderita Ca akibat merokok= A/B sedangkan

Kontribusi merokok terhadap Ca paru-paru = AR

(resiko atribut)= A-B

Menelaah hubungan Kausasi

Secara eksperimental memenuhi kriteria Postulat Robert Koch (utk penyakit menular)

Ada hubungan temporal ada aksi baru terjadi reaksi

Konsisten kesimpulan sama walaupun metode berbeda

Kekuatan asosiasi semakin kuat semakin besar angkanya

Hubungan Dosis-Response

Koherensi

Secara observasional ada bbrp kriteria

Kriteria ini tidak berdiri sendiri lbh baik dpt didukung oleh penjelasan mekanisme penyakit atau proses patologis

Studi Intervensi

Ditambahkan karena tujuan epidemiologi

Studi membuat desain program teoritis yg baik dan diuji secara terbatas

Populasi referensi yg akan jd referensi

Populasi aktual studi dilaksanakan pada populasi yg lbh kecil. Kriteria :

Kesamaan karakteristik demografi

Kemudahan atau aksesivitas

Insidensi penyakit yang akan dicegah

Besarnya populasi diperlukan shg didapat kondisi perbedaan bermakna secara statistik

Seleksi populasi ;

Randomisasi peserta seleksi dan randomisasi peserta

Prosedur double blind tujuan subjektifitas

Stratifikasi kelompok dikelompokkan atas berbagai atribut atau faktor

Intrepretasi hasil penelitian

Kelompok studi/intervensi diuji/ dievaluasi program yg

ditawarkan pada populasi

Kelompok kontrol efek yg terjadi utk melihat signifikasi perbedaan

Penilaian efek

Efek suatu program hrs dilaksanakan double blind tidak ada bias

Efek yg objektif (kuantitatif) biasnya berkurang

Desain sequential

Karena tidak cukup peserta peserta tidak berpartisipasi serentak

Kerugian : bila staf peneliti sudah mengetahui hasil studi kelompok sebelumnya

RINGKASAN PENELITIAN

EPIDEMIOLOGI

Penelitian Epidemiologi

Eksperimental

Observasional

Deskriptif: Mengumpulkan informasi untuk

mengkarakterisasi dan merangkum kondisi atau masalah kesehatan

Analitik: bertumpu pada perbandingan kelompok

untuk menentukan berbagai faktor risiko dalam kontribusinya untuk menyebabkan penyakit

Studi deskriptif: merupakan kategori paling dasar pada

seorang epidemiologis

who/person, where/place, and when/time

Desain penelitian

Cross sectional: seperti survey, penyakit dan

faktor paparan diteliti bersamaan

Cohort atau prospektif: meneliti populasi berdasarkan paparannya kemudian melihat terjadinya penyakit di masa datang RR

Case control atau retrospektif: meneliti kelompok penderita dan melihat kemungkinan faktor paparannya OR

First Incident

In March 1985, a nurse epidemiologist in a county

health department noted, while reviewing surveillance

data, three cases in a single month of hepatitis B of unusual origin. Hepatitis B, or serum hepatitis, is transmitted through sexual contact and by exposure to infected bodily fluids, but these three patients did not

seem to have the usual risk factors. All three people

did, however, indicate having received injections at the same health care facility.

The nurse's immediate questions were: Is this a

coincidence? Did these three cases occur by chance or

is there a link? In this instance, the nurse decided to

pursue an investigation.

Second Incident

At 8:30 in the morning on August 2, 1976, Dr. Robert B. Craven of CDC's Viral Diseases Division received a call from a nurse at a

Veterans' Hospital in Philadelphia, Pennsylvania. The nurse reported

two cases of severe respiratory illness, one of which had been fatal. Both people had attended the annual American Legion Convention held July 21-24 . By the evening of August 2, 71 more of the people attending the convention had the same illness, with symptoms of

acute onset of fever, chills, headache, malaise, dry cough, and

myalgia. Further conversations with local and state public health officials revealed that between July 26 and August 2, 18 conventioneers had died. Deaths were due primarily to pneumonia.

An intense investigation began immediately. The incident became known as the first outbreak of Legionnaires' disease and led to the discovery of the gram-negative pathogen, Legionnella pneumophila.

Third Incident

On October 30, 1989, a New Mexico physician notified the

state's health department of three patients with marked

peripheral eosinophilia and severe myalgia. All three patients had been taking oral preparations of L-tryptophan, a nonprescription drug sold as a dietary supplement in

health food stores. Despite extensive clinical evaluation and

testing, the illness could not be identified.

An investigation followed and resulted in the characterization of eosinophilia-myalgia syndrome, EMS. The investigation implicated a vehicle for exposureL- tryptophan dietary supplementsbefore a suspected agent

was identified, and the product was taken off the market.

Eventually, the problem was traced to a contaminant that had been introduced by changes in the production process at a single manufacturing facility.

Diskusi 1:

Berikut angka

kematian

dihubungkan dengan waktu

Bagaimana menurut Anda, melihat kondisi tersebut?

Diskusi 1: Berikut angka kematian dihubungkan dengan waktu Bagaimana menurut Anda, melihat kondisi tersebut?

Data kematian

diperlihatkan

dalam periode waktu sbb.:

Apa komentar Anda?

Data kematian diperlihatkan dalam periode waktu sbb.: Apa komentar Anda?
Apa komentar Anda mengenai grafik berikut?

Apa komentar

Anda mengenai

grafik berikut?

Latihan 1

You are called to help investigate a cluster of

17 men who developed leukemia in a

community. Some of them worked as electrical repair men, and others were ham

radio operators. Which study design would

you choose to investigate a possible association between exposure to

electromagnetic fields and leukemia?

Latihan 2

The manager of a grocery store has reported a

rash illness among the store’s workers. What

type of study would you use to determine the source of the outbreak? Why? What is the

appropriate measure of association? After

reviewing the table in the Appendix showing

the data on exposure to celery for these

workers, calculate the measure of association

and interpret your results.

Exposed to Celery?

Rash

No Rash

Total

Yes

25

31

56

No

5

65

70

Total

30

96

126

Latihan 1 - Diskusi

Because the total population at risk is not well defined, you

would use a case-control study.

You would begin by enrolling the 17 people already identified with leukemia as the case group.

You would also need to determine what group might serve as an appropriate comparison, or control, group. Neighbors

might be used for the control group, for example.

In your case-control study, you would determine whether each case-patient and each control had been exposed to electromagnetic fields (however you defined that

exposure).

Finally, you would compare the exposures of case-patients and controls.

Latihan 2 Diskusi

You would use a cohort study because the

outbreak is small and confined.

The appropriate measure of association for a cohort study is relative risk, which is calculated in this case as the attack rate for workers exposed to celery divided by the attack rate for those who were not exposed to

celery

The attack rate for exposed workers is 25 / 56, or 44.6%.

The attack rate for workers who were not exposed is 5 / 70,

or 7.1%.

Thus, the relative risk for exposure to celery is 44.6 / 7.1, or

6.3.

This means that workers who were exposed to celery were 6.3 times more likely to develop the rash illness than those who were not exposed, and it is therefore likely that celery was the source of the outbreak.

However, before you could draw this conclusion, you would

need to compare the relative risk for celery with that for

other vegetables and fruits to see if the implication is stronger for any of them.

Then, to test the likelihood of your findings,

you would need to calculate a test of

statistical significance such as chi-square for the item with the highest relative risk and look

up the corresponding p-value in a table of p-

values. If the p-value was below .05, your findings would be considered statistically

significant.

Cohort studies

A cohort study is the best technique for analyzing an

outbreak in a small, well-defined population.

For example, you would use a cohort study if an outbreak of gastroenteritis occurred among people who attended a social function, such as a wedding, and

a complete list of wedding guests was available.

In this situation, you would ask each attendee the same set of questions about potential exposures (e.g., what foods and beverages he or she had consumed at

the wedding) and whether he or she had become ill

with gastroenteritis.

After collecting this information from each guests, you

would be able to calculate an attack rate for people who

ate a particular item (were exposed) and an attack rate for those who did not eat that item (were not exposed).

For the exposed group, the attack rate is found by dividing the number of people who ate the item and became ill by

the total number of people who ate that item.

For those who were not exposed, the attack rate is found by dividing the number of people who did not eat the item but still became ill by the total number of people who did not eat that item.

To identify the source of the outbreak from this

information, you would look for an item with:

a high attack rate among those exposed and

a low attack rate among those not exposed (so

the difference or ratio between attack rates for

the two exposure groups is high); in addition

most of the people who became ill should have

consumed the item, so that the exposure could

explain most, if not all, of the cases.

Usually, you would also calculate the

mathematical association between exposure

(consuming the food or beverage item) and illness for each food and beverage.

This is called the relative risk and is produced

by dividing the attack rate for people who were exposed to the item by the attack rate

for those who were not exposed.

The table on the next page is based on a famous outbreak of gastroenteritis following a church supper in Oswego, New York, in 1940

and illustrates the use of a cohort study.

Number of people who ate specified item

Number of people who did not eat specified item

Food

Ill

Baked Ham

Spinach

Mashed

potatoes*

Cabbage

salad

Jell-O

Rolls

Brown

bread

Milk

Coffee

Water

Cakes

Ice Cream

(van)

Ice Cream

(choc)*

Fruit Salad

Well

Total

Attack

Rate %

Ill

Well

Total

Attack Rate %

29

17

46

63

17

12

29

59

26

17

43

60

20

12

32

62

23

14

37

62

23

14

37

62

18

10

28

64

28

19

47

60

16

7

23

70

30

22

52

58

21

16

37

57

25

13

38

66

18

9

27

67

28

20

48

58

2

2

4

50

44

27

71

62

19

12

31

61

27

17

44

61

13

11

24

54

33

18

51

65

27

13

40

67

19

16

35

54

43

11

54

80

3

18

21

14

25

22

47

53

20

7

27

74

4

2

6

67

42

27

69

61

Of the 80 people who attended the supper, 75 were

interviewed. Forty-six people met the case definition.

Attack rates for those who did and did not eat each of 14 items are presented in the table.

Did most of the 46 people who met the case definition eat that food item? Is the attack rate low among people who

did not eat that item?

You should have identified vanilla ice cream as the implicated vehicle, or source.

The relative risk is calculated as 80 / 14, or 5.7. This relative

risk indicates that people who ate the vanilla ice cream

were 5.7 times more likely to become ill than were those who did not eat the vanilla ice cream.

Case-control studies

In most outbreaks the population is not well defined,

and so cohort studies are not feasible.

In these instances, you would use the case-control study design. In a case-control study, you ask both case-patients and controls about their exposures.

You then can calculate a simple mathematical measure

of associationcalled an odds ratioto quantify the relationship between exposure and disease.

This method does not prove that a particular exposure caused a disease, but it is very helpful and effective in evaluating possible vehicles of disease.

When you design a case-control study, your first, and

perhaps most important, decision is who the controls

should be.

Conceptually, the controls must not have the disease in question, but should be from the same population as

the case-patients. In other words, they should be

similar to the case-patients except that they do not have the disease.

Common control groups consist of neighbors and

friends of case-patients and people from the same

physician practice or hospital as case-patients.

In general, the more case-patients and controls you

have, the easier it will be to find an association. Often,

however, you are limited because the outbreak is small.

For example, in a hospital, 4 or 5 cases may constitute an outbreak. Fortunately, the number of potential controls will usually be more than you need.

In an outbreak of 50 or more cases, 1 control per case- patient will usually suffice. In smaller outbreaks, you might use 2, 3, or 4 controls per case-patient. More

than 4 controls per case-patient will rarely be worth

your effort.

In a case-control study, you cannot calculate

attack rates because you do not know the total

number of people in the community who were and were not exposed to the source of the disease under study.

Without attack rates, you cannot calculate relative risk; instead, the measure of association you use in a case study is an odds ratio.

When preparing to calculate an odds ratio, it is

helpful to look at your data in a 2×2 table.

For instance, suppose you were investigating an outbreak of hepatitis A in a small town, and you suspected that the source was a favorite restaurant of the townspeople. After questioning case-patients and controls about whether they had eaten at that restaurant, your data might look like this:

Ate at Restaurant A?

Total:

Case Patients

Controls

Total

Yes

a = 30

b = 36

66

No

c = 10

d = 70

80

40

106

146

The odds ratio is calculated as ad/bc.

The odds ratio for Restaurant A is thus 30 × 70 /

36 × 10, or 5.8.

This means that people who ate at Restaurant A were 5.8 times more likely to develop hepatitis A than were people who did not eat there.

Even so, you could not conclude that Restaurant

A was the source without comparing its odds ratio with the odds ratios for other possible

sources. It could be that the source is elsewhere

and that it just so happens that many of the

people who were exposed also ate at Restaurant

A.

Testing statistical significance

The final step in testing your hypothesis is to

determine how likely it is that your study results

could have occurred by chance alone.

In other words, how likely is it that the exposure

your study results point to as the source of the

outbreak was not related to the disease after all?

A test of statistical significance is used to evaluate this likelihood. Statistical significance is a broad

area of study, and this will include only a brief

overview here.

The first step in testing for statistical significance is to

assume that the exposure is not related to disease. This

assumption is known as the null hypothesis.

Next, you compute a measure of association, such as a relative risk or an odds ratio.

These measures are then used in calculating a chi-

square test (the statistical test most commonly used in studying an outbreak) or other statistical test.

Once you have a value for chi-square, you look up its corresponding p-value (or probability value) in a table of chi-squares.

In interpreting p-values, you set in advance a cutoff

point beyond which you will consider that chance is a

factor.

A common cutoff point is .05. When a p-value is below the predetermined cutoff point, the finding is

considered "statistically significant," and you may

reject the null hypothesis in favor of the alternative hypothesis, that is you may conclude that the exposure

is associated with disease.

The smaller the p-value, the stronger the evidence that

your finding is statistically significant