|
5. PERFORMING A MANUAL
AUDIT
a. Sample Size Calculations
For
manual auditors, the time requirements of the chart review process generally
make it impractical to review all active DM patients. Rather than reviewing all records, a random sample of records
may be selected for review.
The number of charts you need to select depends on the number of active
patients in your diabetes register.
Table
1 outlines the minimum number of charts you will need to audit to be reasonably
sure (90% confident) that a 10% difference noted from a previous or subsequent
audit is a real change and not just due to chance.
If, for example, your facility has 1000 active patients with diabetes,
you will need to audit a total of 63 charts.
T able 1 - Sample Size Calculations
Sample size needed to be 90% or 95% certain that the rate
you find is within 10% or within 5% of the true rate, for populations up to
4000.
|
Population
|
╒══
90% Certainty══╕
|
╒═95%
Certainty ═╕ |
|
| (# of
DM Patients)
|
Within
10%
|
Within 5%
|
Within 10%>
|
Within 5%
|
|
<30
|
all
|
all
|
all
|
all
|
|
30
|
21
|
27
|
23
|
28
|
|
40
|
25
|
35
|
28
|
36
|
|
50
|
29
|
42
|
33
|
44
|
|
60
|
32
|
49
|
37
|
52
|
|
70
|
34
|
56
|
40
|
59
|
|
80
|
37
|
62
|
44
|
66
|
|
90
|
39
|
68
|
46
|
73
|
|
100
|
40
|
73
|
49
|
79
|
|
110
|
42
|
78
|
51
|
86
|
|
120
|
43
|
83
|
53
|
91
|
|
130
|
44
|
88
|
55
|
97
|
|
140
|
46
|
92
|
57
|
103
|
|
150
|
47
|
96
|
59
|
108
|
|
160
|
48
|
101
|
60
|
113
|
|
170
|
48
|
104
|
61
|
118
|
|
180
|
49
|
108
|
63
|
123
|
|
190
|
50
|
112
|
64
|
127
|
|
200
|
51
|
115
|
65
|
132
|
|
220
|
52
|
121
|
67
|
140
|
|
240
|
53
|
127
|
69
|
148
|
|
260
|
54
|
133
|
70
|
155
|
|
280
|
54
|
138
|
72
|
162
|
|
300
|
55
|
142
|
73
|
168
|
|
320
|
56
|
147
|
74
|
175
|
|
340
|
56
|
151
|
75
|
180
|
|
360
|
57
|
154
|
76
|
186
|
|
380
|
57
|
158
|
77
|
191
|
|
400
|
58
|
161
|
77
|
196
|
|
420
|
58
|
165
|
78
|
201
|
|
440
|
59
|
168
|
79
|
205
|
|
460
|
59
|
170
|
79
|
209
|
|
480
|
59
|
173
|
80
|
213
|
|
500
|
60
|
176
|
81
|
217
|
|
525
|
60
|
179
|
81
|
222
|
|
550
|
60
|
181
|
82
|
226
|
|
575
|
61
|
184
|
82
|
230
|
|
600
|
61
|
186
|
83
|
234
|
|
650
|
61
|
191
|
84
|
241
|
|
700
|
62
|
195
|
84
|
248
|
|
750
|
62
|
199
|
85
|
254
|
|
800
|
62
|
202
|
86
|
260
|
|
900
|
62
|
208
|
87
|
269
|
|
1000
|
63
|
213
|
88
|
278
|
|
2000
|
65
|
238
|
92
|
322
|
|
3000
|
66
|
248
|
93
|
341
|
| 4000 |
67 |
253 |
94 |
350 |
|
▲
Minimum number of charts recommended |
|
|
|
|
|
|
|
|
|
|
For Statisticians and Geeks Only (you know who you are) |
|
|
The audit sample size calculations in Table 1
use the following method:
|
Sample size = n/(1+(n/population))
|
|
|
|
|
|
|
in which n = Z*Z(P(1-P)/D*D) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Sample size = size of sample randomly selected from the population of
active diabetes patients.
Population = total number of active patients in the diabetes register
at your facility.
P = true proportion of audit factor in the population (since this is
never known exactly, it is taken as 50% (i.e., 0.5) as the most conservative
value).
D = (Maximum) difference between sample mean and population mean
(Table
1 lists within 10% and within 5% which corresponds to a D value of 0.1 and
.05, respectively).
Z = area under normal curve corresponding to the desired
confidence level (see table below)
|
Confidence |
Z |
|
|
|
.90 |
1.645 |
|
|
|
.95 |
1.960 |
|
|
|
.99 |
2.575 |
|
|
|
Reference: Kish (1965) |
|
