WWW-publications from the WHO MONICA Project
May 1998
Kari Kuulasmaa1, Hans-Werner Hense2 and Hanna Tolonen1 for theWHO MONICA Project3
1 MONICA Data Centre, National Public Health Institute, Helsinki, Finland
2 Institute of Epidemiology and Social Medicine, University of Münster,
Münster, Germany
3 Annex: Sites and key personnel of the WHO MONICA
Project
This document includes the main findings of unpublished reports:
Thanks are due to Tuula Virman-Ojanen who collected the data from the Survey Procedures Questionnaires and Alun Evans who commented on the text.
The MONICA Centres are funded predominantly by regional and national governments, research councils, and research charities. Coordination is the responsibility of the World Health Organization (WHO), assisted by local fund raising for congresses and workshops. WHO also supports the MONICA Data Centre (MDC) in Helsinki. Not covered by this general description is the ongoing generous support of the MDC by the National Public Health Institute of Finland, and a contribution to WHO from the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland, USA for support of the MDC. The completion of the MONICA Project is generously assisted through a Concerted Action Grant from the European Community. Likewise appreciated are grants from ASTRA Hässle AB, Sweden, Hoechst AG, Germany, Hoffmann-La Roche AG, Switzerland, the Institut de Recherches Internationales Servier (IRIS), France, and Merck & Co. Inc., New Jersey, USA, to support data analysis and preparation of publications.
The aim of this report is to evaluate the quality of the data on blood pressure in the MONICA surveys (1, 2). The quality assessment consists of:
Where possible, the quality assessment gives results specific to the study populations. However, collection of the data necessary for evaluating the influence of cuff size, time of blood pressure measurement and room temperature were only introduced after the initial surveys had been completed. Therefore, for these aspect of the data quality the results will give general ideas rather than population specific results.
A detailed description of the principles applied in the quality assessment are presented in a related publication (3).
The report considers the Reporting Unit Aggregates (RUA) which are foreseen as potential candidates for units of analyses of the MONICA data. In particular, different combinations of Reporting Units (RU) may be used for cross-sectional and trend analyses if all reporting units of the population were not included in all three or two surveys. The RUAs, their abbreviations and Reporting Units as well as the number of subjects in the surveys are listed in Table 1. Note that the RUAs are combined in such a way that they can be directly used in cross-sectional and trend analyses. RUAs are identified by the abbreviation and a version letter, which together give a unique definition of the Reporting Units included. Altogether 56 RUAs are considered in this report.
In UNK-GLA (one of the RUAs) four surveys instead of the usual two or three were done during the MONICA 10-year period. In the current quality assessment the initial, final and only one of the middle surveys (1992) will be considered.
No selection of subjects according to their age was made for this report. Data for men and women have been combined unless specified differently.
The sources of data for this report are the survey core data in the MONICA Data Centre (MDC), Questionnaires on MONICA Population Survey Procedures (Form VI) completed by the MCCs in 1991 and further checked by them regarding the final survey in 1995, and other relevant information.
In MONICA, blood pressure was to be measured in mmHg. However, in one RUA, YUG-NOS, blood pressure was measured in kPa, the unit of the SI standard. The resolution of the scale in a sphygmomanometer with a mmHg scale is 2 mmHg. In a kPa sphygmomanometer the resolution of the scale is 0.2 kPa which is equal to 1.5 mmHg. The data from YUG-NOS in the MONICA database have been converted to mmHg. However, in the analyses of the terminal digits and the proportion of identical readings for the current quality assessment report the original data in kPa were used.
When random zero sphygmomanometers were used, the uncorrected blood pressure values and the random zero values were transferred to the MDC separately. For most of the analyses of the current report the data were corrected for random zero for the analyses. For calculating the proportion of odd readings (Table 9), proportion of different last digits (Table 10) and terminal digit preference score (Table 11) the original readings uncorrected for random zero were used. There is, however, one exception: In the initial survey of NEZ-AUC a random zero sphygmomanometer was used, but only the corrected values were recorded and therefore only the corrected values are available in the MDC.
When the survey core data were received in MDC, the data were checked routinely for the following constraints:
All violations of these constraints were reported to the MCC for their correction or elucidation. Data values outside the constraint limits were acceptable, but the MCC had to check that the values were not unusual owing to data errors. The MCCs were asked to correct values only if they were incorrect. The current unresolved constraint violations are shown in Appendix 1. There are only few unresolved constraint violations for BEL-GHE (Mid), GER-COT (Mid), GER-ERF (Mid, Fin), GER-HAC (Mid), ITA-BRI (Mid), ROM-BUC (Ini) and RUS-NOC (Fin).
According to MONICA Manual:
The recommended instrument is the random-zero sphygmomanometer, although a simple mercury sphygmomanometer is admissible. In any case the Centres should adhere to the instrument used in earlier surveys.
There is evidence that the random-zero sphygmomanometer gives lower estimates of blood pressure in the population than the simple sphygmomanometer. A systematic investigation (4) has identified several reasons for underestimating the individual's blood pressure using the random-zero sphygmomanometer. However, it was concluded that the underestimation using the random-zero sphygmomanometer, not explained by improper use or maintenance of the device, is only about 1 mmHg.
Instructions for the use of the simple and the random-zero sphygmomanometer were given in the MONICA Manual and during training seminars. There is no way of checking the common reasons for the under-reading of the blood pressure when the random-zero sphygmomanometer is used, but if the measurers are trained properly and the same type of sphygmomanometer is used in repeated surveys, the measurements between the surveys should be comparable.
Results: The type of sphygmomanometer used in each RUA is shown in Table 2. 36 RUAs used the simple sphygmomanometer and 15 used the Hawksley random-zero sphygmomanometer (5). Five RUAs deviated from this general rule:
Conclusions: The type of sphygmomanometer used has an influence in the blood pressure level. However, if the same type of sphygmomanometer is used in all surveys of an RUA, the type should have no influence on the estimates of trends. A change in the type of sphygmomanometer between the surveys (FRA-TOU and SWI-VAF) may have a strong influence in the estimates of blood pressure trends. The type of sphygmomanometer used should be reported in the publications.
It was recommended that the bell of the stethoscope should be used, even if the diaphragm was used in the earlier surveys.
Results: The side of stethoscope used is given in Table 2. The bell was used in all surveys in 16 RUAs, the diaphragm was used in all surveys in 32 RUAs, and 8 RUAs changed from diaphragm or either side to bell or from diaphragm to either side.
Conclusion: The bell is recommended because it gives a sharper sound, but there is no clear evidence of bias from the use of the diaphragm. Therefore, the side of stethoscope is not a major concern for MONICA analyses.
According to MONICA Manual:
The cuff size (bladder-size) should be 12-13 cm wide and sufficiently long to surround at least two thirds of the upper arm.
The MCC should not change the size(s) of a cuff used for blood pressure measurement between the surveys. If several cuff sizes are used, the rule for using each should also be kept unchanged and the information recorded.
It has been observed that narrow cuffs give too high blood pressure values for thick arms and wide cuffs give too low blood pressure values for thin arms (8). For this reason it has been recommended that different sized cuffs be chosen to match arms of different size. If cuff sizes are chosen correctly for different arms, blood pressure measurements should be more accurate than if a single cuff-size is used. For the estimation of trends, however, the possible bias in comparability between populations resulting from differences in choice of cuff size is less important than ensuring comparability of measurements within populations between surveys. Therefore, the MCCs should not change the cuff-sizes or the rules for applying different cuffs between the MONICA surveys.
Table 2 shows the widths of cuffs used in the MCCs. There are differences in the widths of cuffs used in the RUAs, and there are changes of more than 0.5cm in the widths between the surveys in AUS-NEW, AUS-PER, CAN-HAL, CZE-CZE, DEN-GLO, FIN-KUO, FIN-NKA, FIN-TUL, FRA-STR, GER-EGE, GER-ERF and ITA-BRI.
To estimate the effect of the different cuff widths on blood pressure, items CUFF to identify the width of cuff used, and ARM, to record the arm circumference were introduced to the survey data transfer format for the final survey. Assuming that the bias related to the width of cuff and the arm circumference are as tabulated in reference (8), Table 3 shows the bias of the overall population mean of blood pressure for men and women in the final survey in the RUAs for which data are available both for CUFF and ARM. (Details of the estimation of the bias are described in Appendix 2.) Such data are available for 23 of the 41 RUAs of the final survey. Among men the bias varies between -3.7 and 2.0 mmHg for systolic and -2.7 and 1.5 mmHg for diastolic blood pressure. Among women the biases are more negative, and vary between -4.9 and -0.5 mmHg for systolic and -3.5 and -0.2 mmHg for diastolic blood pressure. The lowest and highest values are found respectively in France where a cuff width 14 cm was used and in SWE-NSW where cuffs of width 9, 12 and 15 cm were used. The largest biases to the proportion of subjects with high blood pressure were found in FRA-TOU (-4.8), YUG-NOS (-4.5), FRA-STR (-4.3) and SWE-NSW (+4.0).
To get an idea of the impact of the changes in width of cuff between the surveys, the effect of the known change of cuff width in the RUAs where one cuff was used was calculated assuming that the population distributions of arm circumference did not change between the surveys. The results are shown in Table A.
| RUA | Change of cuff width (cm) | Estimated change (mmHg) in bias of population blood pressure mean | |||
|---|---|---|---|---|---|
| SBP | DBP | ||||
| Men | Women | Men | Women | ||
| FIN-KUO | 13->14 | -1.5 | -1.3 | -1.2 | -1.0 |
| FIN-NKA | 13->14 | -1.6 | -1.4 | -1.2 | -1.1 |
| FIN-TUL | 13->14 | -1.5 | -1.2 | -1.2 | -1.0 |
| GER-EGE | 11->12 | -1.3 | -0.9 | -1.3 | -1.0 |
Conclusion: In cross-sectional comparison of blood pressure mean
values between the RUAs, the contribution of differences in cuff width can be up to 5 mmHg
for systolic and 3.5 mmHg for diastolic blood pressure. As data on arm circumference are
available for 23 RUAs in the final survey only, and we are unaware of the accuracy of the
estimates used for the bias related to cuff-width and arm circumference, correction of
data for the bias is not advised.
For the estimation of trends in blood pressure there should not be a major bias for RUAs
where the widths of cuff and the rules for applying them were kept unchanged. For most of
the other RUAs, it is unlikely that the bias induced by changes in the cuff-sizes is more
than 2 mmHg unless the width of the most commonly used cuff changed by more than 1 cm. In
a number of RUAs, an additional narrow (DEN-GLO) or wide (AUS-NEW, AUS-PER, CZE-CZE,
FRA-STR and ITA-BRI) cuff was adopted between the surveys, but in most of the RUAs where
data on the use of the different cuffs are available, the additional cuffs were used
rarely.
If obesity changed substantially in the population, then in principle even a consistent use of same-sized cuff could introduce a bias in the estimate of blood pressure trend. To investigate the magnitude of the potential bias, the association between BMI and arm circumference was calculated in the MONICA data. In all populations and both sexes a 1 kg/m2 change in BMI corresponds to about a 0.6 cm change in arm circumference, which according to reference (8) corresponds to 0.5-1 mmHg bias in systolic and diastolic blood pressure. As in most populations the 10-year change in BMI is less than 1 kg/m2, and very rarely over 2 kg/m2 , the change in obesity is not an important source of bias of trend in blood pressure.
According to the MONICA Manual, the measurement should be planned to precede any painful or anxiety producing procedures such as blood taking or electrocardiography. Blood pressure measurement always preceded blood taking in 42 RUAs, and it was always after in 7 RUAs (Table 4). The order varied in one RUA, one changed from before to after, one changed from after to before, and the order is not known for 3 RUAs which did the initial survey only. Where blood pressure was measured after blood taking, it was done at least 40 minutes later, except in DEN-GLO (10min), the initial survey of FRA-LIL (20min) and FRA-TOU (15 min). If blood pressure was measured at least 20 minutes after venepuncture, we can probably assume that the impact on blood pressure measurement is low.
Blood pressure was measured with the subject in a sitting posture in all RUAs except in BEL-LUX which employed the recumbent posture (Table 4). Note, however, that in CAN-HAL, where the 1st and 3rd measurements were sent to the MDC, the 2nd measurement was taken in the recumbent posture. In the middle and final survey of DEN-GLO, where the 3rd and 4th measurements were sent to the MDC, the 1st and 2nd measurement was taken in recumbent posture. The change in posture in CAN-HAL and DEN-GLO may have had an influence on the blood pressure value.
There is evidence of about 5 mmHg higher values in the left arm compared with the right arm (4). In 50 RUAs blood pressure was measured on the right arm (Table 4). In BEL-LUX and HUN-PEC it was measured on the left arm and in ICE-ICE it was randomly left or right. The information is not known for 3 RUAs which did the initial survey only. The arm was not changed in any RUA between the surveys. The exception of using the left arm should be stated in the publications of cross-sectional analyses.
The MONICA Manual required at least two blood pressure measurements on every subject, and that the first and second should be sent to the MDC. It is generally found that later measurements give lower values than the first. The reason for the decline is not well understood. It may relate to the order of the measurement as well as to the time difference between the measurements. Therefore, the order and the time difference should not change between the surveys. In MONICA the average blood pressure difference between the two measurements sent to the MDC varies considerably among the RUAs, as is seen from Table 8.
We do not know the time difference between the different measurements, but the number of measurements and an indication of which ones were reported to the MDC are shown in the Table 4. In most RUAs two measurements were made, and in nearly all populations the first two were sent to the MDC. The exceptions were:
In addition, the order of the measurements sent to the MDC is not known for the initial survey of BEL-LUX, ISR-TEL, MLT-MLT and ROM-BUC.
Conclusion: The exceptions of SWE-GOT, CAN-HAL and USA-STA should be stated in the publications of cross-sectional analyses. To avoid bias, only the first measurement should be used for SWE-GOT in trend analyses.
Other aspects of the blood pressure measurement procedures are based on the instructions given during the training of the measurers. Table 5 shows the data reported by the MCCs on whether or not instructions on the following aspects of blood pressure measurement were given during the training of measurers:
Training procedures score (TS) was defined from these seven items:
| TS = | 2 | if instructions were given on 6 or 7 of the items; |
| 1 | if instructions were given on 4 or 5 of the items; | |
| 0 | otherwise. |
All except 2 RUAs achieved score "2" in all surveys, one improved, from score "1" to"2" after the initial survey and one had score "1" also in the final survey. Therefore, according to the self-reports, the blood pressure measurers were well trained.
Table 6 shows self-reported data as to whether the following procedures had been applied in the training and testing and/or certification of the blood pressure measurers:
Testing and certification procedures score (CS) was calculated from four subscores:
CS was defined as:
| CS = | 2 | if the sum of the subitem scores was equal to 6 or more and no single subitem score was 0; |
| 1 | if the sum of the subitem scores was equal to 4 or 5 and no more than one single subitem score was 0; | |
| 0 | otherwise. |
Hearing testing has become more common in the middle and final survey than it was in the initial. Nevertheless it was still missing in 10 of the 41 RUAs in the final survey. The testing and certification procedures scores in the initial and final survey are summarised in Table B.
| CS | Number of RUAs | Proportion of RUAs (%) | ||||
|---|---|---|---|---|---|---|
| Ini | Mid | Fin | Ini | Mid | Fin | |
| 2 | 17 | 12 | 14 | 31 | 28 | 34 |
| 1 | 14 | 17 | 15 | 26 | 40 | 37 |
| 0 | 23 | 14 | 12 | 43 | 33 | 29 |
| Total | 54 | 43 | 41 | 100 | 100 | 100 |
There was much more heterogeneity in certification procedures between the RUAs than in training of measurers. Among the RUAs where many surveys were done, each of the three values of the testing and certification score was nearly uniformly distributed both in the initial and the middle survey, and the score changed only in a few populations. There was slight improvement in the final survey.
Table 7 gives data reported by the MCCs on procedures to control the measurers' performance during the survey. The items considered are:
For the purpose of deriving Quality Control procedures Scores (QCS), it
was assumed that these three items should be checked at least bimonthly. Scores were
allotted as follows:
| QCS = | 2 | if the number of valid items was 2 or 3; |
| 1 | if the number of valid items was 1; | |
| 0 | otherwise. |
The quality control procedures scores in the initial, middle and final surveys are summarised in Table C.
| QCS | Number of RUAs | Proportion of RUAs (%) | ||||
|---|---|---|---|---|---|---|
| Ini | Mid | Fin | Ini | Mid | Fin | |
| 2 | 31 | 27 | 32 | 57 | 63 | 78 |
| 1 | 8 | 6 | 6 | 15 | 14 | 15 |
| 0 | 15 | 10 | 3 | 28 | 23 | 7 |
| Total | 54 | 43 | 41 | 100 | 100 | 100 |
There is a steady improvement in the quality control procedures towards the final survey. This is probably a result from the distribution of the earlier quality assessment reports and the risk factor training seminars which were organized before the final surveys.
As sections 5 and 6 evaluated the quality assurance and control procedures reported by the MCCs, the aim of the current section is to find indicators of the quality of blood pressure measurements solely from survey core data transferred from the MCCs to the MDC. Quality indicator scores were calculated from the following specific quality measures:
Table 8 gives population mean values and standard deviations, as well as the difference between the first and second measurement. The most compelling conclusions from the table are:
The extreme values may be real, reflecting true changes in the population, but may also be indicators of data quality problems. The views of the MCCs on some of these changes are given in Section 11.
Table 9 gives the proportion of subjects with incomplete measurements in all subjects whose blood pressure was measured. A subject's measurement was considered complete only if both first and second measurement of both systolic and diastolic blood pressure were recorded. A proportion of incomplete measurements score (PIM) was defined as:
| PIM = | 2 | if less than 5% of the measurements were incomplete; |
| 1 | if 5% - 10% of the measurements were incomplete; | |
| 0 | if at least 10% of the measurements were incomplete. |
The proportion of incomplete measurements is high in
The MONICA Manual requires that blood pressure values are read to the nearest even digit. Table 9 gives the proportion of odd readings for systolic and diastolic blood pressure. A proportion of odd readings score (POR) was defined as:
| POR = | 2 | if less than 5% of systolic readings were odd and less than 5% of diastolic readings were odd; |
| 1 | if at least 5% of systolic or diastolic readings were odd and less than 10% of systolic and diastolic readings were odd; | |
| 0 | if at least 10% of systolic readings were odd or at least 10% of diastolic readings were odd. |
The proportion of odd readings was particularly high in:
In the initial survey also many other RUAs had more than 10% odd readings, but in the final survey all others except USA-STA had less than 2%.
Theoretically it is expected that the terminal digits of the blood pressure readings are distributed uniformly on 0, 2, 4, 6 and 8. Table 10 gives the distribution of the terminal digits of the blood pressure readings. Table 11 gives a digit preference score (DPS). It was calculated from all single even blood pressure readings using the formula:
DPS = 100 * ( X2 / df * N )1/2
where N is the number of observations, X2 is the chisquarestatistic for the test of homogeneity of the terminal digits, and df are the respective degrees of freedom (i.e. df = 4 because there are 5 possible even terminal digits). The DPS ranges from 0 to 100. It is low for high agreement with the ideal of nonpreference and rises consistently with the loss of such agreement. From the values of DPS, terminal digit preference score (TDP) was defined as:
| TDP = | 2 | if both systolic and diastolic DPS <10; |
| 1 | if at least one of the two DPS was between 10 and 20 and none reaches 20; | |
| 0 | otherwise. |
The terminal digit preference is used here as an indicator of the blood pressure measurement quality. Therefore, for the RUAs where a random zero sphygmomanometer was used, the digit preference analyses were calculated from the original readings uncorrected for random zero. Value "0" for the terminal digit preference score indicates a severe problem with the validity of the data.
The number of populations with different terminal digit preference scores is summarised in Table D.
| TDP | Number of RUAs | Proportion of RUAs (%) | ||||
|---|---|---|---|---|---|---|
| Ini | Mid | Fin | Ini | Mid | Fin | |
| 2 | 22 | 24 | 23 | 41 | 56 | 56 |
| 1 | 18 | 9 | 11 | 33 | 21 | 27 |
| 0 | 14 | 10 | 7 | 26 | 23 | 17 |
| Total | 54 | 43 | 41 | 100 | 100 | 100 |
The proportion of RUAs with score "0" decreased clearly towards the final survey, but still remained relatively high despite the feedback given by the earlier QA reports and the training seminars. The proportion of zero as the terminal digit was particularly high (more than 45%) in AUS-PER (final survey), CZE-CZE (initial and middle survey), FRA-LIL (initial survey), FRA-TOU (initial survey), HUN-BUD (initial survey), HUN-PEC (middle survey), ISR-TEL (initial survey), POL-TAR (initial survey), ROM-BUC (initial survey), RUS-NOC (middle survey) and RUS-NOI (middle survey).
Because of normal fluctuation in blood pressure, repeated measurements in the same individual frequently differ. When the standard sphygmomanometer is used, absence of such variability within individuals suggests inaccuracy in second or subsequent measurements. If a random zero sphygmomanometer is used, the measurer does not know when the two measurement values are identical, and therefore inaccurate measurements do not usually result in a high proportion of identical readings. After correction for random-zero errors, the differences between the first and second measurements of both of systolic BP and diastolic BP were calculated for each participant in the survey. Table 12 gives the proportion of subjects whose first and second blood pressure readings were identical. A proportion of identical results in duplicate blood pressure measurements score (PIR) was defined as:
| PIR = | 2 | if less than 33% of the duplicate systolic as well as diastolic BP measurements were identical; |
| 1 | if at least one proportion of identical results, for either systolic or diastolic BP measurements, was between 33% and 50% and none above 50%; | |
| 0 | otherwise. |
The number of populations with different proportion of identical results in duplicate measurements scores (PIR) is summarised in Table E.
| PIR | Number of RUAs | Proportion of RUAs (%) | ||||
|---|---|---|---|---|---|---|
| Ini | Mid | Fin | Ini | Mid | Fin | |
| 2 | 35 | 32 | 35 | 65 | 74 | 85 |
| 1 | 12 | 8 | 4 | 22 | 19 | 10 |
| 0 | 6 | 3 | 2 | 11 | 7 | 5 |
| Not possible to calculate | 1 | 0 | 0 | 2 | 0 | 0 |
| Total | 54 | 43 | 41 | 100 | 100 | 100 |
Most of the populations had PIR score "2" in all surveys. In four RUAs there was a change from "0" to "2". In most cases the high proportion of identical readings follows a high last digit preference. In three cases PIR was "0" even though the last digit preference score was not "0". They were CHN-BEI (initial survey), RUS-NOI (final survey) and SWI-TIC (middle survey). In CHN-BEI the finding is in accord with the fact that the standard deviation of the difference between the first and second measurement is extremely small (Table8).
In SWI-TIC (middle survey) over 60% of the measurements had identical readings even though a random-zero sphygmomanometer was used. This very unexpected finding together with the fact that the mean difference between the two measurements (Table 8) was exactly zero, both for systolic and diastolic blood pressure, raises the suspicion that the data are fabricated. The MCC has investigated the problem. They have neither found any evidence of fabrication nor any explanation for the strange findings.
To assess consistency in the validity and/or precision of blood pressure measurements throughout the period of each survey, within survey time trend scores (WTT) were calculated as follows:
The observations of each survey were divided into sex-specific quintiles in the order of ascending date of examination. Mean systolic BP and diastolic BP, after random-zero correction, were determined for each quintile. Deviations of quintile-specific means from the overall mean were analyzed for both sexes to identify potential time related changes in the validity and/or precision of systolic and diastolic BP measurements. Control limits (CL) were defined for each of the four sub-analyses (men and women, systolic and diastolic) as prediction intervals around the overall sexspecific means according to the formula:
CL = BP + ( SD / Nq1/2 ) * 2.58
where BP is the sexspecific overall mean of SBP and DBP, SD is the overall standard deviation, Nq is the number of observations per quintile, and 2.58 is the 99.5% percentile of the standard normal distribution. The 99.5% percentile was chosen to account for the multiple testing situation (five times). Introduction of the 99.95% percentile, i.e. 3.29 instead of 2.58, defined the "alarm limit". Note that the control limits and the alarm limits depend on the overall standard deviation and therefore on the sample size of each survey. Quintilespecific averages located outside the control and alarm limits indicate unexplained variations of blood pressure results over time provided that there were no time related differences in population sampling procedures.
The four subanalyses of time trends (males SBP [WTT1] and DBP [WTT2], females SBP [WTT3] and DBP [WTT4]) were evaluated and a numerical score was assigned to each within survey time trend as:
|
|
|
|
|
|
|
|
A within survey time trend summary score (WTTS) was calculated summarizing the findings in the four subanalyses. WTTS was defined as:
| WTTS = | 2 | if the sum of the four ratings WTT1 to WTT4 equalled 7 or more; |
| 1 | if the sum of four WTTn ratings equalled 4, 5 or 6 and no more than one WTTn = 0; | |
| 0 | otherwise. |
Table 13 gives the within survey time trend scores and the within survey time trend summary scores. The number of populations with different scores is summarised in Table F.
| WTTS | Number of RUAs | Proportion of RUAs (%) | ||||
|---|---|---|---|---|---|---|
| Ini | Mid | Fin | Ini | Mid | Fin | |
| 2 | 22 | 18 | 19 | 41 | 42 | 46 |
| 1 | 10 | 14 | 13 | 18 | 33 | 32 |
| 0 | 22 | 11 | 9 | 41 | 25 | 22 |
| Total | 54 | 43 | 41 | 100 | 100 | 100 |
In addition to changes in the measurement quality, the time trend score may be low if different locations of the reporting unit area, different sexes or different age groups are examined at different periods. Furthermore, seasonal variation in blood pressure may account for some of the time trends detected in some RUAs. No adjustment for such factors was done when the time trend score was calculated. To detect populations where the age distribution of those examined was different at different periods, the mean age of the participants was calculated for each quintile of the date of examination. Table 13 shows the difference in the mean age between the oldest and the youngest quintiles in each population. If the difference exceeds 5 years, the time trend summary score is given in brackets. The table reveals that the age differences are very large in some surveys of some RUAs (GER-BER, GER-ERF, ITA-FRI, LTU-KAU, RUS-MOC, RUS-MOC, RUS-MOI, RUS-NOC and RUS NOI), suggesting little overlap in the survey periods of different age groups.
Table 14 gives an overview of the data based quality assessment scores and a summary score of blood pressure measurement quality (SSQ). The summary score was defined as:
| SSQ = | 2 | if the sum of the five scores valued 8 or more and no individual score was equal to 0; |
| 1 | if the sum of scores was 4 to 7 and no more than two scores were equal to 0; | |
| 0 | otherwise. |
In addition, if the proportion of zero as last digit exceeded 45% the SSQ was coded as "0", regardless of other quality assessment items. If a low time trend score could be explained by age differences between quintiles of the survey period (i.e. if the time trend score is in brackets in Table 13), then the summary score has been calculated assuming a higher time trend score.
For USA-STA the summary score was not assigned because the summary score would depend on the high proportion of odd readings which are a consequence of the semiautomated measurement procedure rather than the quality of the measurements.
The number of populations with different summary scores is summarised in Table G.
| SSQ | Number of RUAs | Proportion of RUAs (%) | ||||
|---|---|---|---|---|---|---|
| Ini | Mid | Fin | Ini | Mid | Fin | |
| 2 | 26 | 26 | 29 | 48 | 60 | 71 |
| 1 | 20 | 12 | 9 | 37 | 28 | 22 |
| 0 | 7 | 4 | 2 | 13 | 9 | 5 |
| Not possible to calculate | 1 | 1 | 1 | 2 | 2 | 2 |
| Total | 54 | 43 | 41 | 100 | 100 | 100 |
There was a steady improvement in the summary score from one survey to the next. Nevertheless, there were two RUAs (where one includes the other) with score "0" even in the final survey. In these two RUAs the score was better in the earlier surveys.
In addition to the actual blood pressure measurement values, the MONICA core data include four items which are also relevant to the quality of blood pressure measurements or the interpretation of the population blood pressure levels and trends. They are:
The availability of data on these items as well as on the cuff width and arm circumference are given in Table 15. The time of year of the surveys, time of day of blood pressure measurement, room temperature as well as another environmental factor, the characteristics of the blood pressure measurers will be investigated in more detail in the rest of this section.
The season has a potential impact on blood pressure. There is indication in the literature that blood pressure is higher in the cold season than in the warm season. The association between blood pressure and the intensity of the season, however, is not well known. It is not clear if climate associates with long-standing BP elevations/ declines or if it creates short-term effects which are counterbalanced in MONICA measurements by similar indoor measurement conditions like room temperature. Furthermore, the seasonal variation is very different between the countries involved in MONICA.
Therefore, the different survey seasons between the RUAs are likely to have an influence to the cross-sectional comparability between the RUAs except perhaps in the cases where the survey examinations were distributed uniformly throughout the year.
For the assessment of trends in blood pressure it is particularly important that each survey within a RUA took place at the same time of the year. The similarity between the survey periods was assessed in the report "Age, date of examination and survey periods" (9), where "month difference" was used as a measure of the lack of overlap between the surveys. The month difference does not take into account the actual seasonal fluctuation of blood pressure in the population. However, under certain assumptions concerning the seasonal fluctuation of blood pressure, the month difference provides simple upper limits for the seasonal bias in blood pressure trends.
Let us assume:
Then the bias in the risk factor difference between the two surveys is between 0 and XY.
The same result is valid if we replace assumptions a) and b) by
For systolic blood pressure the literature reports seasonal variation between 0 and about 5 mmHg. Therefore, a reasonable upper limit for 6X might be 6 mmHg , i.e. X = 1 mmHg. For X = 1 mmHg we get:
| month difference(Y) | => | maximum bias(XY) |
|---|---|---|
| 1 month | 1 mmHg | |
| 2 months | 2 mmHg | |
| 3 months | 3 mmHg | |
| 4 months | 4 mmHg |
In conclusion, the bias related to a month difference of about two months is very likely less than 2 mmHg, and may be much less depending of the true seasonal fluctuation of blood pressure and the actual survey periods. A month difference of about 2 months is found in:
A month difference of three or four months may already be a source of serious bias. This concerns
For GER-EGE and GER-KMS the month difference is not known.
The mean and standard deviation of the hour of blood pressure measurements in the middle and the final survey are given in Table 16. Both the time of day and the daily duration of the measurements vary substantially between the RUAs. Table 16 also shows the shift in hours between the middle and the final survey, calculated in a way similar to the month distance between the survey periods described in Appendix 2 of reference (9). The largest shift among the 20 RUAs for which it could be calculated was 1.6 hours. Therefore, there is no indication of the shifts in the times of day of blood pressure measurements being a problem for the trend analyses.
The average room temperature during blood pressure measurement is given in Table 17. Among the RUAs for which data are available, the lowest mean value is 20.0 °C and the highest is 23.9 °C. The maximum differences in the average temperature was 3.2 °C, but otherwise always 1.5 °C or less. As the variation in room temperature is small, it is unlikely that room temperature will be a significant source of bias for cross-sectional or trend analyses.
Because blood pressure responses may vary with characteristics of measurers such as age, gender or profession (e.g. white coat hypertension), the measurers should be as alike as possible between the surveys. Table 18 gives the number of measurers and the number in each sex, three age classes and in three professional groups. Mostly there are only small changes in these characteristics between the surveys within the RUAs. In some there was an increase in the age of the measurers. This may be explained by the fact that some MCCs used the same measurers in all three surveys, which actually may strengthen the comparability of the surveys in these RUAs.
To summarise the quality of blood pressure data quality for trend analysis, a score consisting of the SSQ for blood pressure measurement data and other potential sources of bias in the two or three surveys was developed. The score gets a high value for the RUAs with evidence of good quality and a lower value for the RUAs where such evidence is inconsistent or lacking.
The component scores of the summary score are:
| A. | Total SSQ = | 2 | if SSQ=2 in each survey or SSQ=1 in one survey and 2 in the others; |
| 1 | if Total SSQ is not 2 or 0; | ||
| 0 | if SSQ=0 in one of the surveys. |
| B. | Change in device = | 2 | if no change in the type of sphygmomanometer used; |
| 0 | if change. |
| C. | Change in cuff = | 2 | if change not more than 1cm or if the change concerns a cuff which was applied very infrequently; |
| 1 | if change more than 1cm, but applied infrequently; | ||
| 0 | if change more than 1cm applied frequently. |
| D. | Change in posture = | 2 | if no change in the posture of the subject; |
| 0 | if a change. |
| E. | Change in order = | 2 | if no change in the order of the blood pressure measurements used for MONICA; |
| 1 | if change. |
| F. | Change in survey months = | 2 | if month difference < 2 months in all age/sex groups and overall; |
| 1 | if not 2 or 0; | ||
| 0 | if there is an at least 3 month shift for age group 25-64 or 35-64. |
The Quality Score for Trends (QST) is defined using the component scores A-F as:
QST = max{0, min{2A, D, (A+B+C+E+F-6)/2}}
i.e.
| QST = | 0 if A=0 or D=0; |
| max{0, (A+B+C+E+F-6)/2} if A>0 and D>0. |
Table 19 shows the values of the QST for the RUAs for the initial and the final survey and for all three surveys. The results are summarised in Table H.
| Surveys | Ini-Fin | Total | |||||
|---|---|---|---|---|---|---|---|
| Ini-Mid-Fin | QST | 2 | 1.5 | 1 | 0.5 | 0 | |
| 2 | 13 | 0 | 0 | 0 | 0 | 13 | |
| 1.5 | 2 | 7 | 0 | 0 | 0 | 9 | |
| 1 | 2 | 1 | 1 | 0 | 0 | 4 | |
| 0.5 | 0 | 0 | 2 | 1 | 0 | 3 | |
| 0 | 0 | 0 | 2 | 0 | 4 | 6 | |
| No middle survey | 1 | 2 | 0 | 0 | 1 | 4 | |
| Total | 18 | 10 | 5 | 1 | 5 | 39 | |
The score for trend analysis between the initial and final survey is 0 or 0.5 in 6 of the 39 RUAs. If the middle survey is included in the analysis, the score gets worse for 9 RUAs. For such RUAs calculation of 10-year trends using the initial and final survey instead of all three surveys should be considered.
The most important factors for reliable and comparable data on blood pressure measurement are the selection of equipment used for the analysis, in particular the type of sphygmomanometer and width of the cuff; and the measurement procedures.
Both the simple mercury sphygmomanometer and the random-zero sphygmomanometer have been used widely among the MONICA Centres for blood pressure measurement. Both of the devices have advantages. The advantage of the random-zero sphygmomanometer is that it prevents the possible influence of the knowledge of earlier measurements. The advantage of the simple sphygmomanometer is that its use is simpler and therefore it is impossible to make some of the errors which are common sources of bias when the random-zero sphygmomanometer is used. Many studies have shown that the random zero sphygmomanometer gives lower readings than the standard sphygmomanometer. The difference varies from study to study, but it seems to depend mostly on the way the sphygmomanometer is maintained and used (4). The difference may cause a significant bias to the comparison of the blood pressure levels between two populations, and can be serious in estimating trends within a population if different types of devices were used in different surveys. If an MCC uses the same type of sphygmomanometer in all surveys there should be no problem for the estimation of trends, provided that the measurers have been trained properly. In MONICA the type of sphygmomanometer was changed in two RUAs.
Another important equipment issue is the width of the cuff used for blood pressure measurement. This is again a concern in comparison of blood pressure levels between populations, but estimates of trends in blood pressure will be biased only if cuff sizes or the rules for using different cuff sizes are changed between the surveys. It seems that a change in the cuff-sizes used is a problem in a few RUAs only. Nevertheless, standardisation of the cuff size should receive much more attention in future studies than it did in the early stages of MONICA.
The use of standardized measurement procedures, training of the measurers in their use and the quality control during the survey are key issues in monitoring blood pressure levels in the population. Based on the self-reports, training was done well in all populations. The same was not the case with testing and certification procedures and the quality control during the surveys, although there was a clear improvement in the final survey. This is confirmed by the data based QA and can probably be explained by the feed-back given by the quality assessment of the data after the initial and middle survey.
There are several populations where the self-reported data suggests a better or worse summary score of quality than was observed from the actual data. These discrepancies may be due to different interpretations of the questions asked and/or a possible tendency to report more favourably on compliance with quality control procedures than was the actual case. Therefore, it is concluded that quality assessment based on actual survey data should provide a more accurate guide to the true quality of measurements than the qualitative information reported by MCCs on their adherence to quality control procedures.
If the summary score of blood pressure measurement (SSQ) for a survey is "0", there are major reservations concerning the validity of the measurements, and therefore the use of the data. The same concerns the estimation of trends in blood pressure if the quality score for trends (QST) is 0 or 0.5. The data from the middle survey of SWI-TIC should not be used in analyses because of the very strange unexplained characteristics of the data (see Section 7.5).
In most data analyses the mean of two blood pressure measurements will be used, but what to do with the subjects on whom there is one measurement only? In most RUAs such subjects are rare and the decision unimportant. Therefore, the decision should be based on the RUAs with a large number of subjects with one measurement only: GER-RDM (7% with one measurement only in initial survey) SWE-GOT (100% in Ini), and USA-STA (25% in Ini and 12% in Fin). In GER-RDM , one of the reporting units (RU 26) had one measurement only in the initial survey. No middle or final survey was done in the RUA. In SWE-GOT one measurement only was made in the initial survey but two measurements were made in the later surveys. For USA-STA there is evidence that using the single measurement for the subjects on whom there are no data on two measurements will not bias the results significantly: The blood pressure data were obtained using semi-automatic measurements. The reason for the missing readings in the initial survey is that the disks of the semiautomatic measurements were lost and in the final survey there were mechanical failures of the semiautomated devices. The missing observations are distributed uniformly on all Reporting Units and the average difference between the second and third measurements, which the MCC reported to the MDC instead of the first and second measurement, was less than one mmHg (Table 8).
It is suggested that for the respondents with one measurement only, the single measurement will be used for analysis. For trend analyses, only the first measurement should be used for SWE-GOT even in the middle and final survey.
The following methodological details should be mentioned in publications using the data concerned:
The following list includes only the RUAs with specific findings or exceptional background information relevant for the use of the data.
AUS-NEW
AUS-PER
BEL-LUX
CAN-HAL
CZE-CZE
DEN-GLO
FIN-KUO, FIN-NKA and FIN-TUL
FRA-LIL
FRA-STR
FRA-TOU
GER-BRE
GER-COT
GER-EGE
GER-ERF
GER-HAC
GER-RDM
GER-RHN
HUN-BUD
HUN-PEC
ICE-ICE
ISR-TEL
ITA-BRI
MLT-MLT
NEZ-AUC
POL-TAR
POL-WAR
ROM-BUC
RUS-MOC
RUS-MOI
RUS-NOC
RUS-NOI
SWE-GOT
SWI-TIC
SWI-VAF
UNK-GLA
USA-STA
YUG-NOS