Semi-Quantitative Evaluation of Access and Coverage

#> Error in get(paste0(generic, ".", class), envir = get_method_env()) : 
#>   object 'type_sum.accel' not found

In the recent past, measurement of coverage has been mainly through two-stage cluster sampled surveys either as part of a nutrition assessment or through a specific coverage survey known as Centric Systematic Area Sampling (CSAS). However, such methods are resource intensive and often only used for final programme evaluation meaning results arrive too late for programme adaptation. SQUEAC, which stands for Semi-Quantitative Evaluation of Access and Coverage, is a low resource method designed specifically to address this limitation and is used regularly for monitoring, planning and importantly, timely improvement to programme quality, both for agency and Ministry of Health (MoH) led programmes. This package provides functions for use in conducting a SQUEAC investigation.

What does squeacr do?

The squeacr package provides functions that facilitate the processing, analysis and reporting of various components of a SQUEAC investigation. The current version of the squeacr package currently provides the following:

Functions to calculate CMAM programme performance metrics;
Functions to calculate CMAM programme length of stay metrics; and,
Functions to calculate CMAM coverage estimates.

Installation

The squeacr package is not yet available on CRAN but can be installed from the nutriverse R Universe as follows:

install.packages(
  "squeacr",
  repos = c('https://nutriverse.r-universe.dev', 'https://cloud.r-project.org')
)

Usage

Calculating CMAM programme performance metrics

Cure rate, defaulter rate, death rate, and non-response rate are the programme indicators used to monitor performance of CMAM. These indicators are calculated from routine programme monitoring data, an example of which is the monitoring dataset included in squeacr.

CMAM programme monitoring data for Sudan (showing first 15 rows)
State	Locality	Beginning Of Month	New Admissions	Male	Female	Cured	Death	Default	Non-Responder	Total Discharge	Rutf Consumed	Screening	Sites	Month	Year
Gazera	El Qurashi	16	16	8	8	23	0	3	0	26	80	49	NA	Jan	2016
Gazera	El Qurashi	56	24	11	13	0	0	0	0	0	-46	298	NA	Apr	2016
Gazera	El Qurashi	80	41	16	25	22	0	2	0	24	16	225	NA	May	2016
Gazera	El Qurashi	81	43	21	22	29	0	0	0	29	22	215	NA	Jun	2016
Gazera	El Qurashi	93	51	31	30	36	2	0	0	38	14	0	NA	Jul	2016
Gazera	El Qurashi	103	59	34	25	3	0	0	0	3	12	289	NA	Aug	2016
Gazera	El Qurashi	163	69	34	35	8	0	12	2	22	8	0	NA	Sep	2016
Gazera	El Qurashi	104	108	56	40	6	0	47	0	53	-40	0	NA	Oct	2016
Gazera	El Qurashi	275	123	61	62	111	0	81	2	194	32	0	NA	Nov	2016
Gazera	El Qurashi	204	81	39	40	52	0	8	2	62	52	293	NA	Dec	2016
Gazera	El Kamlin	8	8	3	5	0	0	0	0	0	4	8	NA	Jan	2016
Gazera	El Kamlin	119	19	11	8	2	0	2	1	5	16	7	NA	Mar	2016
Gazera	El Kamlin	133	8	5	3	18	0	2	1	21	18	182	NA	Apr	2016
Gazera	El Kamlin	120	22	15	7	8	0	0	1	9	6	552	NA	May	2016
Gazera	El Kamlin	134	9	5	4	15	0	13	0	28	15	285	NA	Jun	2016

The monitoring dataset is from the National CMAM programme in Sudan showing monthly programme statistics per locality. The dataset has the following fields:

Variable	Description
State	Name of state
Locality	Name of locality
Beginning of Month	Cases in programme at beginning of month
New Admissions	New cases admitted within the month
Male	New male cases admitted within the month
Female	New female cases admitted within the month
Cured	Number of cured cases within the month
Death	Number of cases who died within the month
Default	Number of cases who defaulted within the month
Non-Responder	Number of non-responder cases within the month
Total Discharge	Total number of discharges within the month
RUTF Consumed	Number of RUTF consumed
Screening	Screening
Sites	Sites
Month	Month
Year	Year

We can calculate the different programme performance indicators using squeacr. For this example, we’ll calculate the indicators for each state per year.

library(squeacr)
library(dplyr)

monitoring |>
  group_by(State, Year) |>
  summarise(
    total_discharge = sum(`Total Discharge`, na.rm = TRUE),
    cure_rate = calculate_cured(sum(Cured, na.rm = TRUE), total_discharge),
    default_rate = calculate_default(sum(Default, na.rm = TRUE), total_discharge),
    death_rate = calculate_dead(sum(Death, na.rm = TRUE), total_discharge),
    non_response_rate = calculate_no_response(sum(`Non-Responder`, na.rm = TRUE), total_discharge),
    .groups = "drop"
  )

which results in the following:

#> # A tibble: 72 × 7
#>    State          Year  total_discharge cure_rate default_rate death_rate
#>    <chr>          <chr>           <dbl>     <dbl>        <dbl>      <dbl>
#>  1 Blue Nile      2016             9693     0.889       0.0906    0.0151 
#>  2 Blue Nile      2017            10286     0.948       0.0399    0.00972
#>  3 Blue Nile      2018             8807     0.947       0.0404    0.00863
#>  4 Blue Nile      2019             9882     0.953       0.0366    0.00708
#>  5 Central Darfur 2016            13313     0.921       0.0440    0.0174 
#>  6 Central Darfur 2017            18098     0.935       0.0421    0.00912
#>  7 Central Darfur 2018            17600     0.939       0.0364    0.00955
#>  8 Central Darfur 2019            18573     0.952       0.0260    0.00549
#>  9 East Darfur    2016             9895     0.929       0.0550    0.0104 
#> 10 East Darfur    2017            12611     0.956       0.0327    0.00690
#> # ℹ 62 more rows
#> # ℹ 1 more variable: non_response_rate <dbl>

CMAM programme length-of-stay

The length-of-stay in a CMAM programme is an important metric that can provide insight into several aspects of the program’s performance and effectiveness. It is calculated from those discharged cured from outpatient care by counting the number of days between the admission date and the discharge date.

The otp_beneficiaries dataset in the package is an example of a patient record data from which length-of-stay can be calculated using the calculate_los() function:

calculate_los(otp_beneficiaries$admDate, otp_beneficiaries$disDate)

which gives the following results:

#> Warning in calculate_los(otp_beneficiaries$admDate, otp_beneficiaries$disDate):
#> Some admission date/s are not in YYYY-MM-DD format or are not available.
#> Returning NA.
#> Warning in calculate_los(otp_beneficiaries$admDate, otp_beneficiaries$disDate):
#> Some discharge dates are earlier than admisison dates. Returning NA.
#>   [1]  56  42  36  49  42  51  19  75  84  49  90  70  91  20  42  50  14  13
#>  [19]  21  28 107  42  42  77  77  77  31  18  18  11  35  35  14  14  14  14
#>  [37]  28  11  61  73 102  71  71 112  55  71  80  22  22  63  62  44  30  42
#>  [55]  35  35  28  84  28  14  42  34  47  42  45  43  23  42 105 120 105  56
#>  [73] 104  42  79  90  77  28  14  14  77  28  14  54 103  78  79  70  70  98
#>  [91]  78  63  58 125  42  49  44  35  89  86  60  39  41  50  47  46  48  51
#> [109]  50  44  44  46  39  50  54 140  58  84  53  56  21  54  21  28  49  18
#> [127]  56  28  28  21  54  57  29  59  50  39  91 136 127  63  93 155  35 105
#> [145]  42  28  28  35  35  70  35  82  14  17  28 168 147 112  42  35  21  97
#> [163]  35  66  35  28 126  84  70 140  22  63  42  70  94  63  63  98  70  77
#> [181]  77  60  63  63  84  56  49  91  35  42  42  49  70  57  29  64  41  21
#> [199]  93  23  31  28  30  14  21  55  65  28  21  21  88  14  22  21  21  21
#> [217]  35  63  42  28  84  48  14  18  14  14  30  35  81  76  42  28  28  28
#> [235]  56  28  56  42  98  58  35  28  39  34  33  28  49  28  64  28  29  33
#> [253]  80  77  60  42  49  56  55  42  91  98  55  92  98 112  63  63  21  63
#> [271]  63  58  56  63 126  91 119  28  72 111  42  63  91  98  91  84  15  45
#> [289]  NA  29  42  49  42  49  49  14  28  44  35  49  42  84  30  14  14   9
#> [307] 112  56 112  46  28  56  14  70  70  35  28  28  28  48 123  35  14  14
#> [325]  19  14  56  32  35 131  21  47  53  64  64  39  NA  NA  37  32  41   6
#> [343]  42  30  26  44  28  19  15  14  50  35  14  31  28  21   7  26  14  14
#> [361]  28   7   7  19  31  27  20  33  62  28  15  13  28  16  19  30   7  14
#> [379]  36  15   7  43  20 100  64  52  93  34  30  57  NA  56  81  52  95  63
#> [397]  49  54  37  70  84  28  28  66  56

The median length-of-stay in a CMAM programme can be calculated as follows:

calculate_los_median(otp_beneficiaries$admDate, otp_beneficiaries$disDate)

which gives the following results:

#> Warning in calculate_los(admission_date = admission_date, discharge_date =
#> discharge_date): Some admission date/s are not in YYYY-MM-DD format or are not
#> available. Returning NA.
#> Warning in calculate_los(admission_date = admission_date, discharge_date =
#> discharge_date): Some discharge dates are earlier than admisison dates.
#> Returning NA.
#> [1] 43

CMAM programme coverage

The squeacr provides functions to calculate programme coverage. These functions implement the single coverage estimator approach¹. In this approach, treatment coverage is calculated in such a way that estimates the number of severe acute malnutrition (SAM) cases that have not been enrolled in the programme but have been recovering without treatment (r_out).

For example, if a coverage survey yielded 5 SAM cases in the programme, 25 cases not in the programme, and 5 recovering cases in the programme, r_out can be calculated as follows:

calculate_rout(cin = 5, cout = 25, rin = 5)
#> [1] 6

Note here that the calculate_rout() function has another argument named k which is a correction factor representing the ratio of the mean length of an untreated episode to the mean length of a CMAM treatment episode. This, by default, is set to k = 3 in the function. However, this should be adjusted based on programme data to estimate the mean length of a SAM treatment episode.

This calculation for r_out is used within calculate_tc() to estimate treatment coverage:

calculate_tc(cin = 5, cout = 25, rin = 5)
#> [1] 0.2439024