facilities_labels.csv and individuals_labels.csv created in processed_data folderFiO2 analysis
Data cleaning
Merge languages
First task is to combine all the languages into a single file.
no labels
Here is the combined language file, without labels.
labels
And the combined language file with labels.
Split into facility and individual level answers
The next task is to separate the combined language files into two parts, the questions about the facility, and the questions about individual practice.
Remove duplicate facilities
Facility name cleaning
There are a lot of facilities with similar or duplicate names.
Let’s start by listing out all the facility names to see what the patterns are.
there are 170 unique facility namesWe will replace similar names as follows:
# map fat_name to hospital_mapping
facilities['fat_name'] = facilities['fat_name'].map(hospital_mapping)there are 118 unique facility names
we have reduced the number of unique facility names from 170 to 118Duplicate entries by same individual
Exact duplicates
We can easily just get rid of these, if their facility answer is identical and the same name and email.
before dropping duplicates, the number of entries is 204
the number of duplicate entries by same individual is 16
after dropping duplicates, the number of entries is 188Remaining duplicates
No contact information
For the hospitals that still have duplicates, we want to be able to contact someone to verify.
Therefore, we will get rid of duplicate hospitals that have NEITHER a contact email NOR a contact phone number.
after dropping duplicates, the number of entries is 168Drop by number of missing values
- Now we will calculate the number of missing values in each row.
- Then for each hospital, we will drop the rows where the number of missing values is greater than the smallest number of missing values for any hospital.
after filtering, the number of facilities is 155Drop by completed column
Next we will drop those duplicate facilities where the completed column is not 2 (complete)
the number of records in the dataframe is 155
after dropping incomplete facilities, the number of facilities is 130Drop by timestamp
Next, we will find people who responded more than once and had completed surveys, and keep the latest timestamp.
Keeping only facilities with the latest timestamp: 124Drop duplicate records that were not verified
Fred has called the rest of the hospitals, for verification. These are the facilities that were not called, did not get updated information, and therefore can be dropped.
there are 0 duplicate facilitiesFacility Analysis
Describe hospital distribution
count
Nigeria 22
Uganda 19
Tanzania 9
South Africa 6
Namibia 5
Liberia 5
Rwanda 5
Ethiopia 5
Madagascar 3
Burkina Faso 3
Kenya 3
Sudan 2
Zambia 2
Zimbabwe 2
Mozambique 2
Réunion 2
Congo 1
Mayotte 1
Burundi 1
Ghana 1
South Sudan 1
Eswatini 1
Cameroon 1
Name: count, dtype: int64Geocoding
Mapping
Unable to display output for mime type(s): application/vnd.plotly.v1+jsonCategorical variable analysis
What are teh characteristics of the responding facilities?
HTML file created: table.html| Section | Count (proportion) |
|---|---|
| Facility Level | |
| Tertiary | 42 (0.41) |
| Referral | 29 (0.28) |
| General | 24 (0.24) |
| Primary care | 7 (0.07) |
| Hospital funding model | |
| Public | 71 (0.7) |
| Private | 13 (0.13) |
| Public-private partnership | 10 (0.1) |
| Mission (private not-for-profit) | 8 (0.08) |
| Setting or location of hospital | |
| Urban | 63 (0.62) |
| Peri-urban | 21 (0.21) |
| Rural | 18 (0.18) |
| Type of hospital | |
| Academic (Affiliated with a training institution) | 71 (0.7) |
| Non-academic | 31 (0.3) |
| Anesthesia provider cateogry | |
| Specialist physician anesthesia provider | 75 (0.74) |
| Non-physician anesthesia provider (with formal training) | 24 (0.24) |
| Non-specialist physician anesthesia provider like a surgeon or medical officer | 2 (0.02) |
| Providers with no formal training (trained on the job) | 1 (0.01) |
Numeric columns
| count | mean | std | min | 25% | 50% | 75% | max | |
|---|---|---|---|---|---|---|---|---|
| num_ana_mach | 102.0 | 5.176471 | 4.106049 | 0.0 | 2.0 | 4.0 | 8.0 | 20.0 |
| fio2_analys | 102.0 | 3.137255 | 3.653397 | 0.0 | 1.0 | 2.0 | 4.0 | 20.0 |
| num_oxymet | 102.0 | 7.215686 | 10.484904 | 0.0 | 3.0 | 5.0 | 8.0 | 100.0 |
| num_21 | 102.0 | 3.235294 | 3.661890 | 0.0 | 1.0 | 2.0 | 5.0 | 20.0 |
| num_fio2_var | 102.0 | 3.392157 | 3.912804 | 0.0 | 1.0 | 2.0 | 4.0 | 20.0 |
| num_100 | 102.0 | 4.705882 | 4.255662 | 0.0 | 1.0 | 3.5 | 8.0 | 20.0 |
| num_peep | 102.0 | 4.107843 | 3.891872 | 0.0 | 1.0 | 3.0 | 5.0 | 20.0 |
Histograms:
Select all that apply
This is the count and percentage of ‘checked’ answers for each source of oxygen.
| Count | Percentage ("Checked") | Source | |
|---|---|---|---|
| 13 | 43 | 0.421569 | No access to medical air |
| 18 | 42 | 0.411765 | All our machines can provide FiO2 ranges betwe... |
| 17 | 42 | 0.411765 | FiO2 analyzer is unavailable or dysfunctional |
| 1 | 36 | 0.352941 | Piping from the manifold |
| 4 | 34 | 0.333333 | Cylinders (externally purchased) |
| 0 | 32 | 0.313725 | Piping from plant |
| 15 | 24 | 0.235294 | Supply of medical gases |
| 8 | 20 | 0.196078 | Piping from the manifold |
| 3 | 18 | 0.176471 | Cylinders (filled at hospital plant) |
| 7 | 16 | 0.156863 | Piping from the plant |
| 16 | 14 | 0.137255 | Missing components like hoses |
| 5 | 13 | 0.127451 | Anesthesia machine in-built concentrator |
| 12 | 10 | 0.098039 | Anesthesia machine entrains air |
| 2 | 10 | 0.098039 | Stand-alone concentrator |
| 10 | 9 | 0.088235 | Cylinders (externally purchased) |
| 14 | 6 | 0.058824 | Purity of medical gases |
| 9 | 6 | 0.058824 | Stand-alone air compressor |
| 11 | 3 | 0.029412 | Cylinders (filled at the hospital plant) |
| 6 | 1 | 0.009804 | No access to oxygen |
Bar plots of counts, and stacked bars including ‘checked’ and ‘unchecked’
Individual analysis
Cleaning
there are 170 unique facility names
there are 118 unique facility names
we have reduced the number of unique facility names from 148 to 118
the number of records in the dataframe is 204
after dropping incomplete facilities, the number of facilities is 152
Keeping only facilities with the latest timestamp: 145before dropping duplicates, the number of entries is 145
the number of duplicate entries by same individual is 0
after dropping duplicates, the number of entries is 102GA with ETT
GA with ETT Count Percentage ("Checked")
0 Yes 133 0.917241
1 No 12 0.082759
Abdominal surgeries per week
count 129.000000
mean 16.598450
std 16.847301
min 0.200000
25% 5.000000
50% 12.000000
75% 20.000000
max 100.000000
Name: abd_sur_wkly, dtype: float64
carrier gases
None Count Percentage ("Checked")
0 Pure oxygen alone 71 0.550388
1 Oxygen and medical air 48 0.372093
2 Oxygen, Nitrous oxide, and Medical air 6 0.046512
3 Oxygen and nitrous oxide 4 0.031008<Figure size 1000x600 with 0 Axes>
Other frequency questions
Anesthesia provider work experience
count 129.000000
mean 5.752171
std 5.387062
min 0.200000
25% 2.000000
50% 4.000000
75% 7.000000
max 29.000000
Name: anes_prov_exper, dtype: float64
Hindrances
| Count | Percentage ("Checked") | Source | |
|---|---|---|---|
| 0 | 64 | 0.441379 | Lack of FiO2 analyzer |
| 1 | 30 | 0.206897 | Limited training (knowledge and skill) |
| 2 | 58 | 0.400000 | Other limitations of the anesthesia machine |
| 3 | 55 | 0.379310 | Access to medical gases (availability and cost) |
| 4 | 6 | 0.041379 | Hospital and or national practice policies and... |
| 5 | 43 | 0.296552 | I am able to use variable ranges of FiO2 at my... |
Sex
sex
2.0 89
1.0 40
Name: count, dtype: int64Age
count 129.000000
mean 37.937984
std 7.764857
min 20.000000
25% 34.000000
50% 37.000000
75% 42.000000
max 69.000000
Name: age, dtype: float64