Underestimation of consumables costs

[sakshimohan]

Currently we are underestimating the cost of consumables by $69 million or 38% of the 2018 expenditure reported in the Resource Mapping data.

This has occurred after merging in PR #1510 which updates the logging of consumables. This is likely because - i. we're not logging enough quantities of consumables Used for certain HSIs, ii. certain consumables are not included in the model at all. In order to expose this issue, I have compared -

1. The estimated consumables cost to expenditure (2018) and projected budget data (2020-22) from Resource Mapping Round 7, and
2. Logged Items_Used data to quantities dispensed reported in the OpenLMIS 2018 - The challenge with this is that we do not always know the units used in the OpenLMIS data so the focus here is to expose individual consumables which were dispensed as per OpenLMIS but not recorded as used in the TLO model.

The .csv file attached below summarises some of the key consumables for which we might be underestimating the quantity of consumables needed.
pending_consumables_issues_4Dec2024.csv
The columns to focus on are -

1. Quantity dispensed as per OpenLMIS (unit unknown)
2. Quantity dispensed as per TLO model (see chosen unit for TLO model)
3. Module lead - to find which issues are addressed to you
4. Notes - In this column, I describe the issue in detail and provide context for how comparison with resource mapping data reveals a discrepancy.

This figure summarises the issue. Detailed table comparison TLO cost estimates with resource mapping data below.

[sakshimohan]

Current status of calibration (Areas that need addressing in //bold//)

Cost category	Calibration category	Actual Expenditure (RM) 2018	Maximum (Annual Projected Budget 2020-2022)	TLO Cost Estimate ($)	Expenditure as per RM minus TLO Cost estimate
Facility operating costs	Facility utility bills	12,142,707.98	31,595,108.55	25,731,776.58
Facility operating costs	Infrastructure - Rehabilitation	4,117,616.05	11,684,723.49	5,570,953.60
Facility operating costs	Vehicles - Fuel and Maintenance	24,393,757.96	32,411,573.39	18,710,184.48
Human resources for health	Health Worker Salaries	114,418,952.79	133,228,884.25	107,230,088.91
Human resources for health	Health Worker Training - In-Service	12,897,136.55	17,743,710.12	12,086,818.36
Human resources for health	Health Worker Training - Pre-Service	1,277,697.67	2,665,199.15	2,852,164.56
Human resources for health	Mentorships & Supportive Supervision	55,553,332.80	66,652,165.02	52,062,955.21
Medical consumables	//Antimalarials//	681,848.59	8,094,056.65	25,609.00	656,239.59
Medical consumables	Antiretrovirals	53,796,360.12	77,506,903.49	53,787,113.67
Medical consumables	//Bednets//	3,552,600.00	28,061,308.64	463,707.63	3,088,892.37
Medical consumables	//Cervical Cancer//	534,872.32	1,581,977.74	101,750.65	433,121.67
Medical consumables	Condoms and Lubricants	210,313.84	11,211,211.72	423,034.86
Medical consumables	//HIV Screening/Diagnostic Tests//	34,469,044.13	94,345,358.80	21,024,664.01	13,444,380.12
Medical consumables	//Indoor Residual Spray//	7,272,001.18	13,194,497.82
Medical consumables	//Malaria RDTs//	1,719,000.00	5,643,435.92	538,367.86	1,180,632.14
Medical consumables	//Other Drugs, medical supplies, and commodities//	39,437,013.14	80,801,692.06	32,532,025.40
Medical consumables	//Other family planning commodities//	12,323,143.36	21,052,691.19	1,981,220.53	10,341,922.83
Medical consumables	Supply Chain	21,341,274.59	37,668,917.17	9,696,423.30
Medical consumables	//TB Tests (including RDTs) //	599,011.62	6,695,563.77	53,719.77	545,291.85
Medical consumables	//TB Treatment//	2,338,536.34	4,685,956.00	1,664,427.22	674,109.12
Medical consumables	//Undernutrition commodities//	913,214.58	1,721,063.60	23,268.04	889,946.54
Medical consumables	Vaccines	4,200,418.52	13,209,399.57	9,319,174.55
Medical consumables	//Voluntary Male Medical Circumcision//	2,891,681.00	1,832,645.01	94,754.61	2,796,926.39
Medical equipment	//Medical Equipment - Maintenance//	2,294,331.38	16,468,941.94	712,772.18
Medical equipment	//Medical Equipment - Purchase//	3,754,149.89	19,024,579.29	525,327.86

[sakshimohan]

Thanks @joehcollins and @EvaJanouskova or getting on the call just now. I'm adding the full list of consumables recorded as used in the model here - consumables_used_full_list_6dec2024.csv

[EvaJanouskova]

@sakshimohan,

The undernutrition commodities may need to be looked up again, once the wasting module is in as the food supplements are there, and I can also add the Vit A (not included at the moment). @tdm32, is there Vit A logged with active measles cases (the guideline for management of acute malnutrition says it should be)?

Regarding the hormonal methods expenditure, I'll need to look at this more closely. Will get back to you later on that.

I think that Zinc and ORS sachet are related to diarrhoea module which is in @joehcollins' hands.

[sakshimohan]

wasting module is in as the food supplements are there, and I can also add the Vit A (not included at the moment). @tdm32, is there Vit A logged with ac

Thanks @EvaJanouskova - That sounds good to wait for the wasting module before I check nutrition commodities again.

[tdm32]

@sakshimohan,

The undernutrition commodities may need to be looked up again, once the wasting module is in as the food supplements are there, and I can also add the Vit A (not included at the moment). @tdm32, is there Vit A logged with active measles cases (the guideline for management of acute malnutrition says it should be)?

Regarding the hormonal methods expenditure, I'll need to look at this more closely. Will get back to you later on that.

I think that Zinc and ORS sachet are related to diarrhoea module which is in @joehcollins' hands.

Hi @EvaJanouskova yes Vit A is given to all with diagnosed measles infection

[joehcollins]

@sakshimohan do you have the details on the simulation you used for these results - mainly population size?

[sakshimohan]

@sakshimohan do you have the details on the simulation you used for these results - mainly population size?

Hi @joehcollins The population size for the year I've used for validation (2018) is 18,253,723. This is the breakdown of the female population by age if helpful -
age_range
0-4 1.433140e+06
10-14 1.205246e+06
100+ 0.000000e+00
15-19 1.034697e+06
20-24 8.792392e+05
25-29 7.499530e+05
30-34 6.022888e+05
35-39 4.632901e+05
40-44 3.668343e+05
45-49 2.865175e+05
5-9 1.311676e+06
50-54 2.307145e+05
55-59 1.822685e+05
60-64 1.414413e+05
65-69 1.134090e+05
70-74 9.439114e+04
75-79 6.359625e+04
80-84 3.541849e+04
85-89 1.654608e+04
90-94 6.513745e+03
95-99 1.017773e+03

[joehcollins]

sorry @sakshimohan i meant modelled individuals

[sakshimohan]

sorry @sakshimohan i meant modelled individuals

Ah got it. The modelled population was 125,544.8 in 2018. The simulation started with 100,000.
And female population by age group -
age_range
0-4 9856.8
10-14 8289.4
100+ 0.0
15-19 7116.4
20-24 6047.2
25-29 5158.0
30-34 4142.4
35-39 3186.4
40-44 2523.0
45-49 1970.6
5-9 9021.4
50-54 1586.8
55-59 1253.6
60-64 972.8
65-69 780.0
70-74 649.2
75-79 437.4
80-84 243.6
85-89 113.8
90-94 44.8
95-99 7.0

[sakshimohan]

Adding a summary of consumable availability by level and item code for reference here - consumable_availability_summary.csv

[joehcollins]

@sakshimohan here are my initial thoughts on the consumables assigned to me. I think there are some actions we could take but might be good to meet and discuss this first?:

Vincristin 1mg Injection:

• Indicated (STG 2023) as treatment for lymphoma/retinoblastoma which is not explicitly modelled
• To discuss with Andrew - we could add as a treatment for other cancers but would be difficult to ensure the correct amount is distributed...

Cyclophosphamide, 1 g

• Cyclophosphomide is indicated as treatment for breast cancer in Malawi STG hence why we are dispensing here - unclear why there is no distribution in Malawi according to LMIS
• Would need to speak to an oncologist to know more?

bleomycin powder 15 mg (15,000 IU) ()_1_IDA**

• Indicated (STG 2023) as treatment for lymphoma which is not explicitly modelled.
• We could add as a treatment for other cancers but would be difficult to ensure the correct amount is distributed (as above)

Catheter Foley's + urine bag (2000ml) 14g_1_CMST

• There are other items which are the same/similar to this item code that are called (i.e. foley catheter etc). Can we match these??

Bandage, plaster of paris 15cm_12_CMST

• There are other items which are the same/similar to this item code that are called. Can we match theses??

Ampicillin injection 250mg, PFR_1_CMST

• There are other items which are the same/similar to this item code that are called. Can we match theses??

Benzylpenicillin 1g (1MU), PFR_1_CMST, Benzylpenicillin 3g (5MU), PFR_1_CMST

• Large number of indications for this drug - many of which we do not model explicitly. Currently we’re only modelling for RMNCH. Unsure best way to resolve?

Gentamicin Sulphate 40mg/ml, 2ml_1_CMST

• Large number of indications for this drug - many of which we do not model explicitly. Currently we’re only modelling for RMNCH. Unsure best way to resolve?

Flucloxacillin 250mg_100_CMST

• As per the code in master this should be used in ALRI. I imagine its not being used because the indication for use in the model is rare (2nd_line_Antibiotic_therapy_for_severe_staph_pneumonia)?
• In addition (similarly with the antibiotics above) the number of clinical indications for this drug are huge. We are not capturing all of them

Prednisolone 5mg_100_CMST

• Prednisolone is defined as a consumable in COPD but never called...
• In addition (similarly with the antibiotics) the number of clinical indications for this drug are huge. We are not capturing them

Salbutamol, tablet, 4 mg, Salbutamol, syrup, 2 mg/5 ml, Salbutamol sulphate 1mg/ml, 5ml_1_CMST

• Inhaled salbutamol is captured through ALRI module.
• We don’t model asthma for which oral salbutamol (for children) would be the leading indication – maybe could be given in wheeze for ALRI instead of inhaler

Tetracycline, tablet, 500 mg

• Similar issue to other antibiotics in the model

Epinephrine, ampoule, 1 mg/ml

• I don’t think we sufficiently model the indication for this (anaphylaxis, cardiac arrest, vasoconstriction in ICU)

Ketamine, 10 ml vial, 50 mg/ml

• I think this would be indicated in RTI but isn’t currently modelled. Similarly there may be some RMCH indications but I would need to think on those.

Erythromycin, tablet, 250 mg

• Huge number of indications. Not currently modelled.

Insulin soluble 100 IU/ml, 10ml_1_CMST

• As per the code in master this should be used in pregnancy for gestational diabetes. I imagine its not being used because the indication for use in the model is rare?
• Also should be modelled for diabetes?

Glibenclamide 5mg_1000_CMST

• As per the code in master this should be used in pregnancy for gestational diabetes. I imagine its not being used because the indication for use in the model is rare?
• Also should be modelled for diabetes?

vitamin B12 (cyanocobalamine) 1 mg/ml, 1 ml, inj._100_IDA

• We don’t model treatment for Vitamin b12 anaemia

Plaster of Paris (POP) 10cm x 7.5cm slab_12_CMST

• Modelled in RTI. We would need to increase the amount used in the model

[tdm32]

PR #1543 addresses HIV test under-use

[tdm32]

PR Tara malaria cons fix dec2024 #1544 addresses the issues with malaria consumables noted here. We can have some further discussion on ITN logging and these changes (if needed) can be added to this PR also.

[joehcollins]

@sakshimohan just summarising our call:

1.) Consumables to check if they are ever delivered in a model run - Flucloxacillin 250mg_100_CMST, Insulin soluble 100 IU/ml, 10ml_1_CMST, Glibenclamide 5mg_1000_CMST, Prednisolone 5mg_100_CMST

2.) Consumables which could be matched with other consumables that are being delivered - Bandage, plaster of paris 15cm_12_CMST, Ampicillin injection 250mg, PFR_1_CMST, Catheter Foley's + urine bag (2000ml) 14g_1_CMST

[EvaJanouskova]

@sakshimohan, Btw, are the costs generated by TLO for the same year, 2018?

[sakshimohan]

@sakshimohan, Btw, are the costs generated by TLO for the same year, 2018?

Yes Eva. That's right.

[EvaJanouskova]

@sakshimohan,

Regarding the hormonal methods expenditure, I'll need to look at this more closely. Will get back to you later on that.

I think this is because we do not distinguish between essential and non-essential items, which is causing the contraceptives not to be issued as often as they should be. (This would solve the issue #975.)

[EvaJanouskova]

Hi @sakshimohan , Could you please post here the job ID and the username of the person who send the job so I could look up what are the proportions of women on contraceptives?

[sakshimohan]

Hi @sakshimohan , Could you please post here the job ID and the username of the person who send the job so I could look up what are the proportions of women on contraceptives?

Hi @EvaJanouskova. This is based on draw 0 from the following job ID and username -

tlo batch-download --username t.mangal@imperial.ac.uk hss_elements-2024-11-12T172311Z

[EvaJanouskova]

To get the proportions of women on contraceptives, I need to run this with contraception logging.

@tdm32, is it the scenario: src/scripts/comparison_of_horizontal_and_vertical_programs/scenario_hss_elements.py? And I need to run 1 draw and baseline scenario only, right?

[tdm32]

To get the proportions of women on contraceptives, I need to run this with contraception logging.

@tdm32, is it the scenario: src/scripts/comparison_of_horizontal_and_vertical_programs/scenario_hss_elements.py? And I need to run 1 draw and baseline scenario only, right?

Hi @EvaJanouskova if you want the proportion of women on contraception in a baseline scenario with no interventions etc. you could run any analysis script, just keeping everything as default.